apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-06 17:17:14 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Modify YOLOv7 Infer file and the external model namespace (#87)

Browse Source 
* first commit for yolov7

* pybind for yolov7

* CPP README.md

* CPP README.md

* modified yolov7.cc

* README.md

* python file modify

* delete license in fastdeploy/

* repush the conflict part

* README.md modified

* README.md modified

* file path modified

* file path modified

* file path modified

* file path modified

* file path modified

* README modified

* README modified

* move some helpers to private

* add examples for yolov7

* api.md modified

* api.md modified

* api.md modified

* YOLOv7

* yolov7 release link

* yolov7 release link

* yolov7 release link

* copyright

* change some helpers to private

* change variables to const and fix documents.

* gitignore

* Transfer some funtions to private member of class

* Transfer some funtions to private member of class

* Merge from develop (#9)

* Fix compile problem in different python version (#26)

* fix some usage problem in linux

* Fix compile problem

Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>

* Add PaddleDetetion/PPYOLOE model support (#22)

* add ppdet/ppyoloe

* Add demo code and documents

* add convert processor to vision (#27)

* update .gitignore

* Added checking for cmake include dir

* fixed missing trt_backend option bug when init from trt

* remove un-need data layout and add pre-check for dtype

* changed RGB2BRG to BGR2RGB in ppcls model

* add model_zoo yolov6 c++/python demo

* fixed CMakeLists.txt typos

* update yolov6 cpp/README.md

* add yolox c++/pybind and model_zoo demo

* move some helpers to private

* fixed CMakeLists.txt typos

* add normalize with alpha and beta

* add version notes for yolov5/yolov6/yolox

* add copyright to yolov5.cc

* revert normalize

* fixed some bugs in yolox

* fixed examples/CMakeLists.txt to avoid conflicts

* add convert processor to vision

* format examples/CMakeLists summary

* Fix bug while the inference result is empty with YOLOv5 (#29)

* Add multi-label function for yolov5

* Update README.md

Update doc

* Update fastdeploy_runtime.cc

fix variable option.trt_max_shape wrong name

* Update runtime_option.md

Update resnet model dynamic shape setting name from images to x

* Fix bug when inference result boxes are empty

* Delete detection.py

Co-authored-by: Jason <jiangjiajun@baidu.com>
Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>
Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com>
Co-authored-by: huangjianhui <852142024@qq.com>

* first commit for yolor

* for merge

* Develop (#11)

* Fix compile problem in different python version (#26)

* fix some usage problem in linux

* Fix compile problem

Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>

* Add PaddleDetetion/PPYOLOE model support (#22)

* add ppdet/ppyoloe

* Add demo code and documents

* add convert processor to vision (#27)

* update .gitignore

* Added checking for cmake include dir

* fixed missing trt_backend option bug when init from trt

* remove un-need data layout and add pre-check for dtype

* changed RGB2BRG to BGR2RGB in ppcls model

* add model_zoo yolov6 c++/python demo

* fixed CMakeLists.txt typos

* update yolov6 cpp/README.md

* add yolox c++/pybind and model_zoo demo

* move some helpers to private

* fixed CMakeLists.txt typos

* add normalize with alpha and beta

* add version notes for yolov5/yolov6/yolox

* add copyright to yolov5.cc

* revert normalize

* fixed some bugs in yolox

* fixed examples/CMakeLists.txt to avoid conflicts

* add convert processor to vision

* format examples/CMakeLists summary

* Fix bug while the inference result is empty with YOLOv5 (#29)

* Add multi-label function for yolov5

* Update README.md

Update doc

* Update fastdeploy_runtime.cc

fix variable option.trt_max_shape wrong name

* Update runtime_option.md

Update resnet model dynamic shape setting name from images to x

* Fix bug when inference result boxes are empty

* Delete detection.py

Co-authored-by: Jason <jiangjiajun@baidu.com>
Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>
Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com>
Co-authored-by: huangjianhui <852142024@qq.com>

* Yolor (#16)

* Develop (#11) (#12)

* Fix compile problem in different python version (#26)

* fix some usage problem in linux

* Fix compile problem

Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>

* Add PaddleDetetion/PPYOLOE model support (#22)

* add ppdet/ppyoloe

* Add demo code and documents

* add convert processor to vision (#27)

* update .gitignore

* Added checking for cmake include dir

* fixed missing trt_backend option bug when init from trt

* remove un-need data layout and add pre-check for dtype

* changed RGB2BRG to BGR2RGB in ppcls model

* add model_zoo yolov6 c++/python demo

* fixed CMakeLists.txt typos

* update yolov6 cpp/README.md

* add yolox c++/pybind and model_zoo demo

* move some helpers to private

* fixed CMakeLists.txt typos

* add normalize with alpha and beta

* add version notes for yolov5/yolov6/yolox

* add copyright to yolov5.cc

* revert normalize

* fixed some bugs in yolox

* fixed examples/CMakeLists.txt to avoid conflicts

* add convert processor to vision

* format examples/CMakeLists summary

* Fix bug while the inference result is empty with YOLOv5 (#29)

* Add multi-label function for yolov5

* Update README.md

Update doc

* Update fastdeploy_runtime.cc

fix variable option.trt_max_shape wrong name

* Update runtime_option.md

Update resnet model dynamic shape setting name from images to x

* Fix bug when inference result boxes are empty

* Delete detection.py

Co-authored-by: Jason <jiangjiajun@baidu.com>
Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>
Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com>
Co-authored-by: huangjianhui <852142024@qq.com>

Co-authored-by: Jason <jiangjiajun@baidu.com>
Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>
Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com>
Co-authored-by: huangjianhui <852142024@qq.com>

* Develop (#13)

* Fix compile problem in different python version (#26)

* fix some usage problem in linux

* Fix compile problem

Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>

* Add PaddleDetetion/PPYOLOE model support (#22)

* add ppdet/ppyoloe

* Add demo code and documents

* add convert processor to vision (#27)

* update .gitignore

* Added checking for cmake include dir

* fixed missing trt_backend option bug when init from trt

* remove un-need data layout and add pre-check for dtype

* changed RGB2BRG to BGR2RGB in ppcls model

* add model_zoo yolov6 c++/python demo

* fixed CMakeLists.txt typos

* update yolov6 cpp/README.md

* add yolox c++/pybind and model_zoo demo

* move some helpers to private

* fixed CMakeLists.txt typos

* add normalize with alpha and beta

* add version notes for yolov5/yolov6/yolox

* add copyright to yolov5.cc

* revert normalize

* fixed some bugs in yolox

* fixed examples/CMakeLists.txt to avoid conflicts

* add convert processor to vision

* format examples/CMakeLists summary

* Fix bug while the inference result is empty with YOLOv5 (#29)

* Add multi-label function for yolov5

* Update README.md

Update doc

* Update fastdeploy_runtime.cc

fix variable option.trt_max_shape wrong name

* Update runtime_option.md

Update resnet model dynamic shape setting name from images to x

* Fix bug when inference result boxes are empty

* Delete detection.py

Co-authored-by: Jason <jiangjiajun@baidu.com>
Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>
Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com>
Co-authored-by: huangjianhui <852142024@qq.com>

* documents

* documents

* documents

* documents

* documents

* documents

* documents

* documents

* documents

* documents

* documents

* documents

* Develop (#14)

* Fix compile problem in different python version (#26)

* fix some usage problem in linux

* Fix compile problem

Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>

* Add PaddleDetetion/PPYOLOE model support (#22)

* add ppdet/ppyoloe

* Add demo code and documents

* add convert processor to vision (#27)

* update .gitignore

* Added checking for cmake include dir

* fixed missing trt_backend option bug when init from trt

* remove un-need data layout and add pre-check for dtype

* changed RGB2BRG to BGR2RGB in ppcls model

* add model_zoo yolov6 c++/python demo

* fixed CMakeLists.txt typos

* update yolov6 cpp/README.md

* add yolox c++/pybind and model_zoo demo

* move some helpers to private

* fixed CMakeLists.txt typos

* add normalize with alpha and beta

* add version notes for yolov5/yolov6/yolox

* add copyright to yolov5.cc

* revert normalize

* fixed some bugs in yolox

* fixed examples/CMakeLists.txt to avoid conflicts

* add convert processor to vision

* format examples/CMakeLists summary

* Fix bug while the inference result is empty with YOLOv5 (#29)

* Add multi-label function for yolov5

* Update README.md

Update doc

* Update fastdeploy_runtime.cc

fix variable option.trt_max_shape wrong name

* Update runtime_option.md

Update resnet model dynamic shape setting name from images to x

* Fix bug when inference result boxes are empty

* Delete detection.py

Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>
Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com>
Co-authored-by: huangjianhui <852142024@qq.com>

Co-authored-by: Jason <jiangjiajun@baidu.com>
Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>
Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com>
Co-authored-by: huangjianhui <852142024@qq.com>
Co-authored-by: Jason <928090362@qq.com>

* add is_dynamic for YOLO series (#22)

* modify namespace for yolov5lite

* modify namespace for external detection models

* modify namespace for yolov5lite

* modify namespace for face models

* py namespace for external model

* fix compile error and infer file

* change paddleyolox to yolox

* delete space in cmakelist

* check code specification

Co-authored-by: Jason <jiangjiajun@baidu.com>
Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>
Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com>
Co-authored-by: huangjianhui <852142024@qq.com>
Co-authored-by: Jason <928090362@qq.com>
This commit is contained in:
ziqi-jin
2022-08-10 10:28:07 +08:00
committed by GitHub
parent de2a87a8f2
commit 9918374d74
88 changed files with 1460 additions and 1404 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
csrcs/fastdeploy/vision.h
View File

@@ -15,29 +15,27 @@
#include "fastdeploy/core/config.h"
#ifdef ENABLE_VISION
#include "fastdeploy/vision/biubug6/retinaface.h"
#include "fastdeploy/vision/deepcam/yolov5face.h"
#include "fastdeploy/vision/deepinsight/arcface.h"
#include "fastdeploy/vision/deepinsight/cosface.h"
#include "fastdeploy/vision/deepinsight/insightface_rec.h"
#include "fastdeploy/vision/deepinsight/partial_fc.h"
#include "fastdeploy/vision/deepinsight/scrfd.h"
#include "fastdeploy/vision/deepinsight/vpl.h"
#include "fastdeploy/vision/detection/contrib/nanodet_plus.h"
#include "fastdeploy/vision/detection/contrib/scaledyolov4.h"
#include "fastdeploy/vision/detection/contrib/yolor.h"
#include "fastdeploy/vision/detection/contrib/yolov5.h"
#include "fastdeploy/vision/detection/contrib/yolov5lite.h"
#include "fastdeploy/vision/detection/contrib/yolov6.h"
#include "fastdeploy/vision/detection/contrib/yolov7.h"
#include "fastdeploy/vision/detection/ppdet/model.h"
#include "fastdeploy/vision/linzaer/ultraface.h"
#include "fastdeploy/vision/megvii/yolox.h"
#include "fastdeploy/vision/meituan/yolov6.h"
#include "fastdeploy/vision/detection/contrib/yolox.h"
#include "fastdeploy/vision/facedet/contrib/retinaface.h"
#include "fastdeploy/vision/facedet/contrib/scrfd.h"
#include "fastdeploy/vision/facedet/contrib/ultraface.h"
#include "fastdeploy/vision/facedet/contrib/yolov5face.h"
#include "fastdeploy/vision/faceid/contrib/arcface.h"
#include "fastdeploy/vision/faceid/contrib/cosface.h"
#include "fastdeploy/vision/faceid/contrib/insightface_rec.h"
#include "fastdeploy/vision/faceid/contrib/partial_fc.h"
#include "fastdeploy/vision/faceid/contrib/vpl.h"
#include "fastdeploy/vision/matting/contrib/modnet.h"
#include "fastdeploy/vision/ppcls/model.h"
#include "fastdeploy/vision/ppogg/yolov5lite.h"
#include "fastdeploy/vision/detection/ppdet/model.h"
#include "fastdeploy/vision/ppseg/model.h"
#include "fastdeploy/vision/rangilyu/nanodet_plus.h"
#include "fastdeploy/vision/ultralytics/yolov5.h"
#include "fastdeploy/vision/zhkkke/modnet.h"
#endif
#include "fastdeploy/vision/visualize/visualize.h"

144
csrcs/fastdeploy/vision/deepinsight/deepinsight_pybind.cc
View File

@@ -1,144 +0,0 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindDeepInsight(pybind11::module& m) {
auto deepinsight_module =
m.def_submodule("deepinsight", "https://github.com/deepinsight");
// Bind SCRFD
pybind11::class_<vision::deepinsight::SCRFD, FastDeployModel>(
deepinsight_module, "SCRFD")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::deepinsight::SCRFD& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::FaceDetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::deepinsight::SCRFD::size)
.def_readwrite("padding_value",
&vision::deepinsight::SCRFD::padding_value)
.def_readwrite("is_mini_pad", &vision::deepinsight::SCRFD::is_mini_pad)
.def_readwrite("is_no_pad", &vision::deepinsight::SCRFD::is_no_pad)
.def_readwrite("is_scale_up", &vision::deepinsight::SCRFD::is_scale_up)
.def_readwrite("stride", &vision::deepinsight::SCRFD::stride)
.def_readwrite("use_kps", &vision::deepinsight::SCRFD::use_kps)
.def_readwrite("max_nms", &vision::deepinsight::SCRFD::max_nms)
.def_readwrite("downsample_strides",
&vision::deepinsight::SCRFD::downsample_strides)
.def_readwrite("num_anchors", &vision::deepinsight::SCRFD::num_anchors)
.def_readwrite("landmarks_per_face",
&vision::deepinsight::SCRFD::landmarks_per_face);
// Bind InsightFaceRecognitionModel
pybind11::class_<vision::deepinsight::InsightFaceRecognitionModel,
FastDeployModel>(deepinsight_module,
"InsightFaceRecognitionModel")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::deepinsight::InsightFaceRecognitionModel& self,
pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size",
&vision::deepinsight::InsightFaceRecognitionModel::size)
.def_readwrite("alpha",
&vision::deepinsight::InsightFaceRecognitionModel::alpha)
.def_readwrite("beta",
&vision::deepinsight::InsightFaceRecognitionModel::beta)
.def_readwrite("swap_rb",
&vision::deepinsight::InsightFaceRecognitionModel::swap_rb)
.def_readwrite(
"l2_normalize",
&vision::deepinsight::InsightFaceRecognitionModel::l2_normalize);
// Bind ArcFace
pybind11::class_<vision::deepinsight::ArcFace,
vision::deepinsight::InsightFaceRecognitionModel>(
deepinsight_module, "ArcFace")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::deepinsight::ArcFace& self, pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::deepinsight::ArcFace::size)
.def_readwrite("alpha", &vision::deepinsight::ArcFace::alpha)
.def_readwrite("beta", &vision::deepinsight::ArcFace::beta)
.def_readwrite("swap_rb", &vision::deepinsight::ArcFace::swap_rb)
.def_readwrite("l2_normalize",
&vision::deepinsight::ArcFace::l2_normalize);
// Bind CosFace
pybind11::class_<vision::deepinsight::CosFace,
vision::deepinsight::InsightFaceRecognitionModel>(
deepinsight_module, "CosFace")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::deepinsight::CosFace& self, pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::deepinsight::CosFace::size)
.def_readwrite("alpha", &vision::deepinsight::CosFace::alpha)
.def_readwrite("beta", &vision::deepinsight::CosFace::beta)
.def_readwrite("swap_rb", &vision::deepinsight::CosFace::swap_rb)
.def_readwrite("l2_normalize",
&vision::deepinsight::CosFace::l2_normalize);
// Bind Partial FC
pybind11::class_<vision::deepinsight::PartialFC,
vision::deepinsight::InsightFaceRecognitionModel>(
deepinsight_module, "PartialFC")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::deepinsight::PartialFC& self, pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::deepinsight::PartialFC::size)
.def_readwrite("alpha", &vision::deepinsight::PartialFC::alpha)
.def_readwrite("beta", &vision::deepinsight::PartialFC::beta)
.def_readwrite("swap_rb", &vision::deepinsight::PartialFC::swap_rb)
.def_readwrite("l2_normalize",
&vision::deepinsight::PartialFC::l2_normalize);
// Bind VPL
pybind11::class_<vision::deepinsight::VPL,
vision::deepinsight::InsightFaceRecognitionModel>(
deepinsight_module, "VPL")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::deepinsight::VPL& self, pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::deepinsight::VPL::size)
.def_readwrite("alpha", &vision::deepinsight::VPL::alpha)
.def_readwrite("beta", &vision::deepinsight::VPL::beta)
.def_readwrite("swap_rb", &vision::deepinsight::VPL::swap_rb)
.def_readwrite("l2_normalize", &vision::deepinsight::VPL::l2_normalize);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/rangilyu/nanodet_plus.cc → csrcs/fastdeploy/vision/detection/contrib/nanodet_plus.cc
View File

@@ -10,7 +10,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/rangilyu/nanodet_plus.h"
#include "fastdeploy/vision/detection/contrib/nanodet_plus.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -18,7 +18,7 @@ namespace fastdeploy {
namespace vision {
namespace rangilyu {
namespace detection {
struct NanoDetPlusCenterPoint {
int grid0;
@@ -350,6 +350,6 @@ bool NanoDetPlus::Predict(cv::Mat* im, DetectionResult* result,
return true;
}
} // namespace rangilyu
} // namespace detection
} // namespace vision
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/rangilyu/nanodet_plus.h → csrcs/fastdeploy/vision/detection/contrib/nanodet_plus.h
View File

@@ -22,7 +22,7 @@ namespace fastdeploy {
namespace vision {
namespace rangilyu {
namespace detection {
class FASTDEPLOY_DECL NanoDetPlus : public FastDeployModel {
public:
@@ -34,7 +34,7 @@ class FASTDEPLOY_DECL NanoDetPlus : public FastDeployModel {
const Frontend& model_format = Frontend::ONNX);
// 定义模型的名称
std::string ModelName() const { return "RangiLyu/nanodet"; }
std::string ModelName() const { return "nanodet"; }
// 模型预测接口，即用户调用的接口
// im 为用户的输入数据，目前对于CV均定义为cv::Mat
@@ -96,6 +96,6 @@ class FASTDEPLOY_DECL NanoDetPlus : public FastDeployModel {
bool is_dynamic_input_;
};
} // namespace rangilyu
} // namespace detection
} // namespace vision
} // namespace fastdeploy

22
csrcs/fastdeploy/vision/megvii/megvii_pybind.cc → csrcs/fastdeploy/vision/detection/contrib/nanodet_plus_pybind.cc
View File

@@ -15,27 +15,25 @@
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindMegvii(pybind11::module& m) {
auto megvii_module =
m.def_submodule("megvii", "https://github.com/megvii/YOLOX");
pybind11::class_<vision::megvii::YOLOX, FastDeployModel>(
megvii_module, "YOLOX")
void BindNanoDetPlus(pybind11::module& m) {
pybind11::class_<vision::detection::NanoDetPlus, FastDeployModel>(
m, "NanoDetPlus")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::megvii::YOLOX& self, pybind11::array& data,
[](vision::detection::NanoDetPlus& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::DetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::megvii::YOLOX::size)
.def_readwrite("size", &vision::detection::NanoDetPlus::size)
.def_readwrite("padding_value",
&vision::megvii::YOLOX::padding_value)
.def_readwrite("is_decode_exported",
&vision::megvii::YOLOX::is_decode_exported)
&vision::detection::NanoDetPlus::padding_value)
.def_readwrite("keep_ratio", &vision::detection::NanoDetPlus::keep_ratio)
.def_readwrite("downsample_strides",
&vision::megvii::YOLOX::downsample_strides)
.def_readwrite("max_wh", &vision::megvii::YOLOX::max_wh);
&vision::detection::NanoDetPlus::downsample_strides)
.def_readwrite("max_wh", &vision::detection::NanoDetPlus::max_wh)
.def_readwrite("reg_max", &vision::detection::NanoDetPlus::reg_max);
}
} // namespace fastdeploy

12
csrcs/fastdeploy/vision/ultralytics/yolov5.cc → csrcs/fastdeploy/vision/detection/contrib/yolov5.cc
View File

@@ -12,17 +12,17 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/ultralytics/yolov5.h"
#include "fastdeploy/vision/detection/contrib/yolov5.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
namespace fastdeploy {
namespace vision {
namespace ultralytics {
namespace detection {
void LetterBox(Mat* mat, std::vector<int> size, std::vector<float> color,
bool _auto, bool scale_fill = false, bool scale_up = true,
int stride = 32) {
void YOLOv5::LetterBox(Mat* mat, std::vector<int> size,
std::vector<float> color, bool _auto, bool scale_fill,
bool scale_up, int stride) {
float scale =
std::min(size[1] * 1.0 / mat->Height(), size[0] * 1.0 / mat->Width());
if (!scale_up) {
@@ -290,6 +290,6 @@ bool YOLOv5::Predict(cv::Mat* im, DetectionResult* result, float conf_threshold,
return true;
}
} // namespace ultralytics
} // namespace detection
} // namespace vision
} // namespace fastdeploy

25
csrcs/fastdeploy/vision/ultralytics/yolov5.h → csrcs/fastdeploy/vision/detection/contrib/yolov5.h
View File

@@ -19,7 +19,7 @@
namespace fastdeploy {
namespace vision {
namespace ultralytics {
namespace detection {
class FASTDEPLOY_DECL YOLOv5 : public FastDeployModel {
public:
@@ -30,7 +30,7 @@ class FASTDEPLOY_DECL YOLOv5 : public FastDeployModel {
const Frontend& model_format = Frontend::ONNX);
// 定义模型的名称
std::string ModelName() const { return "ultralytics/yolov5"; }
std::string ModelName() const { return "yolov5"; }
// 模型预测接口，即用户调用的接口
// im 为用户的输入数据，目前对于CV均定义为cv::Mat
@@ -81,21 +81,28 @@ class FASTDEPLOY_DECL YOLOv5 : public FastDeployModel {
// conf_threshold 后处理时过滤box的置信度阈值
// nms_iou_threshold 后处理时NMS设定的iou阈值
// multi_label 后处理时box选取是否采用多标签方式
bool Postprocess(
FDTensor& infer_result, DetectionResult* result,
bool Postprocess(FDTensor& infer_result, DetectionResult* result,
const std::map<std::string, std::array<float, 2>>& im_info,
float conf_threshold, float nms_iou_threshold, bool multi_label);
float conf_threshold, float nms_iou_threshold,
bool multi_label);
// 查看输入是否为动态维度的 不建议直接使用 不同模型的逻辑可能不一致
bool IsDynamicInput() const { return is_dynamic_input_; }
// whether to inference with dynamic shape (e.g ONNX export with dynamic shape or not.)
// YOLOv5 official 'export_onnx.py' script will export dynamic ONNX by default.
// while is_dynamic_shape if 'false', is_mini_pad will force 'false'. This value will
void LetterBox(Mat* mat, std::vector<int> size, std::vector<float> color,
bool _auto, bool scale_fill = false, bool scale_up = true,
int stride = 32);
// whether to inference with dynamic shape (e.g ONNX export with dynamic shape
// or not.)
// YOLOv5 official 'export_onnx.py' script will export dynamic ONNX by
// default.
// while is_dynamic_shape if 'false', is_mini_pad will force 'false'. This
// value will
// auto check by fastdeploy after the internal Runtime already initialized.
bool is_dynamic_input_;
};
} // namespace ultralytics
} // namespace detection
} // namespace vision
} // namespace fastdeploy

25
csrcs/fastdeploy/vision/meituan/meituan_pybind.cc → csrcs/fastdeploy/vision/detection/contrib/yolov5_pybind.cc
View File

@@ -15,27 +15,24 @@
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindMeituan(pybind11::module& m) {
auto meituan_module =
m.def_submodule("meituan", "https://github.com/meituan/YOLOv6");
pybind11::class_<vision::meituan::YOLOv6, FastDeployModel>(
meituan_module, "YOLOv6")
void BindYOLOv5(pybind11::module& m) {
pybind11::class_<vision::detection::YOLOv5, FastDeployModel>(m, "YOLOv5")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::meituan::YOLOv6& self, pybind11::array& data,
[](vision::detection::YOLOv5& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::DetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::meituan::YOLOv6::size)
.def_readwrite("padding_value",
&vision::meituan::YOLOv6::padding_value)
.def_readwrite("is_mini_pad", &vision::meituan::YOLOv6::is_mini_pad)
.def_readwrite("is_no_pad", &vision::meituan::YOLOv6::is_no_pad)
.def_readwrite("is_scale_up", &vision::meituan::YOLOv6::is_scale_up)
.def_readwrite("stride", &vision::meituan::YOLOv6::stride)
.def_readwrite("max_wh", &vision::meituan::YOLOv6::max_wh);
.def_readwrite("size", &vision::detection::YOLOv5::size)
.def_readwrite("padding_value", &vision::detection::YOLOv5::padding_value)
.def_readwrite("is_mini_pad", &vision::detection::YOLOv5::is_mini_pad)
.def_readwrite("is_no_pad", &vision::detection::YOLOv5::is_no_pad)
.def_readwrite("is_scale_up", &vision::detection::YOLOv5::is_scale_up)
.def_readwrite("stride", &vision::detection::YOLOv5::stride)
.def_readwrite("max_wh", &vision::detection::YOLOv5::max_wh)
.def_readwrite("multi_label", &vision::detection::YOLOv5::multi_label);
}
} // namespace fastdeploy

7
csrcs/fastdeploy/vision/ppogg/yolov5lite.cc → csrcs/fastdeploy/vision/detection/contrib/yolov5lite.cc
View File

@@ -12,13 +12,13 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/ppogg/yolov5lite.h"
#include "fastdeploy/vision/detection/contrib/yolov5lite.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
namespace fastdeploy {
namespace vision {
namespace ppogg {
namespace detection {
void YOLOv5Lite::LetterBox(Mat* mat, const std::vector<int>& size,
const std::vector<float>& color, bool _auto,
@@ -191,7 +191,6 @@ bool YOLOv5Lite::PostprocessWithDecode(
std::vector<YOLOv5Lite::Anchor> anchors;
int num_anchors = anchor_config[0].size() / 2;
GenerateAnchors(size, downsample_strides, &anchors, num_anchors);
// infer_result shape might look like (1,n,85=5+80)
float* data = static_cast<float*>(infer_result.Data());
for (size_t i = 0; i < infer_result.shape[1]; ++i) {
@@ -395,6 +394,6 @@ bool YOLOv5Lite::Predict(cv::Mat* im, DetectionResult* result,
return true;
}
} // namespace ppogg
} // namespace detection
} // namespace vision
} // namespace fastdeploy

7
csrcs/fastdeploy/vision/ppogg/yolov5lite.h → csrcs/fastdeploy/vision/detection/contrib/yolov5lite.h
View File

@@ -19,7 +19,7 @@
namespace fastdeploy {
namespace vision {
namespace ppogg {
namespace detection {
class FASTDEPLOY_DECL YOLOv5Lite : public FastDeployModel {
public:
@@ -30,8 +30,7 @@ class FASTDEPLOY_DECL YOLOv5Lite : public FastDeployModel {
const Frontend& model_format = Frontend::ONNX);
// 定义模型的名称
virtual std::string ModelName() const { return "ppogg/YOLOv5-Lite"; }
virtual std::string ModelName() const { return "YOLOv5-Lite"; }
// 模型预测接口，即用户调用的接口
// im 为用户的输入数据，目前对于CV均定义为cv::Mat
// result 为模型预测的输出结构体
@@ -134,6 +133,6 @@ class FASTDEPLOY_DECL YOLOv5Lite : public FastDeployModel {
// auto check by fastdeploy after the internal Runtime already initialized.
bool is_dynamic_input_;
};
} // namespace ppogg
} // namespace detection
} // namespace vision
} // namespace fastdeploy

28
csrcs/fastdeploy/vision/ppogg/ppogg_pybind.cc → csrcs/fastdeploy/vision/detection/contrib/yolov5lite_pybind.cc
View File

@@ -15,29 +15,29 @@
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindPpogg(pybind11::module& m) {
auto ppogg_module =
m.def_submodule("ppogg", "https://github.com/ppogg/YOLOv5-Lite");
pybind11::class_<vision::ppogg::YOLOv5Lite, FastDeployModel>(ppogg_module,
void BindYOLOv5Lite(pybind11::module& m) {
pybind11::class_<vision::detection::YOLOv5Lite, FastDeployModel>(m,
"YOLOv5Lite")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::ppogg::YOLOv5Lite& self, pybind11::array& data,
[](vision::detection::YOLOv5Lite& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::DetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::ppogg::YOLOv5Lite::size)
.def_readwrite("padding_value", &vision::ppogg::YOLOv5Lite::padding_value)
.def_readwrite("is_mini_pad", &vision::ppogg::YOLOv5Lite::is_mini_pad)
.def_readwrite("is_no_pad", &vision::ppogg::YOLOv5Lite::is_no_pad)
.def_readwrite("is_scale_up", &vision::ppogg::YOLOv5Lite::is_scale_up)
.def_readwrite("stride", &vision::ppogg::YOLOv5Lite::stride)
.def_readwrite("max_wh", &vision::ppogg::YOLOv5Lite::max_wh)
.def_readwrite("anchor_config", &vision::ppogg::YOLOv5Lite::anchor_config)
.def_readwrite("size", &vision::detection::YOLOv5Lite::size)
.def_readwrite("padding_value",
&vision::detection::YOLOv5Lite::padding_value)
.def_readwrite("is_mini_pad", &vision::detection::YOLOv5Lite::is_mini_pad)
.def_readwrite("is_no_pad", &vision::detection::YOLOv5Lite::is_no_pad)
.def_readwrite("is_scale_up", &vision::detection::YOLOv5Lite::is_scale_up)
.def_readwrite("stride", &vision::detection::YOLOv5Lite::stride)
.def_readwrite("max_wh", &vision::detection::YOLOv5Lite::max_wh)
.def_readwrite("anchor_config",
&vision::detection::YOLOv5Lite::anchor_config)
.def_readwrite("is_decode_exported",
&vision::ppogg::YOLOv5Lite::is_decode_exported);
&vision::detection::YOLOv5Lite::is_decode_exported);
}
} // namespace fastdeploy

12
csrcs/fastdeploy/vision/meituan/yolov6.cc → csrcs/fastdeploy/vision/detection/contrib/yolov6.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/meituan/yolov6.h"
#include "fastdeploy/vision/detection/contrib/yolov6.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,11 +20,11 @@ namespace fastdeploy {
namespace vision {
namespace meituan {
namespace detection {
void LetterBox(Mat* mat, std::vector<int> size, std::vector<float> color,
bool _auto, bool scale_fill = false, bool scale_up = true,
int stride = 32) {
void YOLOv6::LetterBox(Mat* mat, std::vector<int> size,
std::vector<float> color, bool _auto, bool scale_fill,
bool scale_up, int stride) {
float scale = std::min(size[1] * 1.0f / static_cast<float>(mat->Height()),
size[0] * 1.0f / static_cast<float>(mat->Width()));
if (!scale_up) {
@@ -262,6 +262,6 @@ bool YOLOv6::Predict(cv::Mat* im, DetectionResult* result, float conf_threshold,
return true;
}
} // namespace meituan
} // namespace detection
} // namespace vision
} // namespace fastdeploy

25
csrcs/fastdeploy/vision/meituan/yolov6.h → csrcs/fastdeploy/vision/detection/contrib/yolov6.h
View File

@@ -22,7 +22,7 @@ namespace fastdeploy {
namespace vision {
namespace meituan {
namespace detection {
class FASTDEPLOY_DECL YOLOv6 : public FastDeployModel {
public:
@@ -33,7 +33,7 @@ class FASTDEPLOY_DECL YOLOv6 : public FastDeployModel {
const Frontend& model_format = Frontend::ONNX);
// 定义模型的名称
std::string ModelName() const { return "meituan/YOLOv6"; }
std::string ModelName() const { return "YOLOv6"; }
// 模型预测接口，即用户调用的接口
// im 为用户的输入数据，目前对于CV均定义为cv::Mat
@@ -61,7 +61,8 @@ class FASTDEPLOY_DECL YOLOv6 : public FastDeployModel {
bool is_scale_up;
// padding stride, for is_mini_pad
int stride;
// for offseting the boxes by classes when using NMS, default 4096 in meituan/YOLOv6
// for offseting the boxes by classes when using NMS, default 4096 in
// meituan/YOLOv6
float max_wh;
private:
@@ -81,21 +82,27 @@ class FASTDEPLOY_DECL YOLOv6 : public FastDeployModel {
// im_info 为预处理记录的信息，后处理用于还原box
// conf_threshold 后处理时过滤box的置信度阈值
// nms_iou_threshold 后处理时NMS设定的iou阈值
bool Postprocess(
FDTensor& infer_result, DetectionResult* result,
bool Postprocess(FDTensor& infer_result, DetectionResult* result,
const std::map<std::string, std::array<float, 2>>& im_info,
float conf_threshold, float nms_iou_threshold);
// 查看输入是否为动态维度的 不建议直接使用 不同模型的逻辑可能不一致
bool IsDynamicInput() const { return is_dynamic_input_; }
// whether to inference with dynamic shape (e.g ONNX export with dynamic shape or not.)
// meituan/YOLOv6 official 'export_onnx.py' script will export static ONNX by default.
// while is_dynamic_input if 'false', is_mini_pad will force 'false'. This value will
void LetterBox(Mat* mat, std::vector<int> size, std::vector<float> color,
bool _auto, bool scale_fill = false, bool scale_up = true,
int stride = 32);
// whether to inference with dynamic shape (e.g ONNX export with dynamic shape
// or not.)
// meituan/YOLOv6 official 'export_onnx.py' script will export static ONNX by
// default.
// while is_dynamic_input if 'false', is_mini_pad will force 'false'. This
// value will
// auto check by fastdeploy after the internal Runtime already initialized.
bool is_dynamic_input_;
};
} // namespace meituan
} // namespace detection
} // namespace vision
} // namespace fastdeploy

37
csrcs/fastdeploy/vision/detection/contrib/yolov6_pybind.cc Normal file
View File

@@ -0,0 +1,37 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindYOLOv6(pybind11::module& m) {
pybind11::class_<vision::detection::YOLOv6, FastDeployModel>(m, "YOLOv6")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::detection::YOLOv6& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::DetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::detection::YOLOv6::size)
.def_readwrite("padding_value", &vision::detection::YOLOv6::padding_value)
.def_readwrite("is_mini_pad", &vision::detection::YOLOv6::is_mini_pad)
.def_readwrite("is_no_pad", &vision::detection::YOLOv6::is_no_pad)
.def_readwrite("is_scale_up", &vision::detection::YOLOv6::is_scale_up)
.def_readwrite("stride", &vision::detection::YOLOv6::stride)
.def_readwrite("max_wh", &vision::detection::YOLOv6::max_wh);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/megvii/yolox.cc → csrcs/fastdeploy/vision/detection/contrib/yolox.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/megvii/yolox.h"
#include "fastdeploy/vision/detection/contrib/yolox.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace megvii {
namespace detection {
struct YOLOXAnchor {
int grid0;
@@ -334,6 +334,6 @@ bool YOLOX::Predict(cv::Mat* im, DetectionResult* result, float conf_threshold,
return true;
}
} // namespace megvii
} // namespace detection
} // namespace vision
} // namespace fastdeploy

18
csrcs/fastdeploy/vision/megvii/yolox.h → csrcs/fastdeploy/vision/detection/contrib/yolox.h
View File

@@ -22,7 +22,7 @@ namespace fastdeploy {
namespace vision {
namespace megvii {
namespace detection {
class FASTDEPLOY_DECL YOLOX : public FastDeployModel {
public:
@@ -33,7 +33,7 @@ class FASTDEPLOY_DECL YOLOX : public FastDeployModel {
const Frontend& model_format = Frontend::ONNX);
// 定义模型的名称
std::string ModelName() const { return "megvii/YOLOX"; }
std::string ModelName() const { return "YOLOX"; }
// 模型预测接口，即用户调用的接口
// im 为用户的输入数据，目前对于CV均定义为cv::Mat
@@ -79,8 +79,7 @@ class FASTDEPLOY_DECL YOLOX : public FastDeployModel {
// im_info 为预处理记录的信息，后处理用于还原box
// conf_threshold 后处理时过滤box的置信度阈值
// nms_iou_threshold 后处理时NMS设定的iou阈值
bool Postprocess(
FDTensor& infer_result, DetectionResult* result,
bool Postprocess(FDTensor& infer_result, DetectionResult* result,
const std::map<std::string, std::array<float, 2>>& im_info,
float conf_threshold, float nms_iou_threshold);
@@ -93,13 +92,16 @@ class FASTDEPLOY_DECL YOLOX : public FastDeployModel {
// 查看输入是否为动态维度的 不建议直接使用 不同模型的逻辑可能不一致
bool IsDynamicInput() const { return is_dynamic_input_; }
// whether to inference with dynamic shape (e.g ONNX export with dynamic shape or not.)
// megvii/YOLOX official 'export_onnx.py' script will export static ONNX by default.
// while is_dynamic_shape if 'false', is_mini_pad will force 'false'. This value will
// whether to inference with dynamic shape (e.g ONNX export with dynamic shape
// or not.)
// megvii/YOLOX official 'export_onnx.py' script will export static ONNX by
// default.
// while is_dynamic_shape if 'false', is_mini_pad will force 'false'. This
// value will
// auto check by fastdeploy after the internal Runtime already initialized.
bool is_dynamic_input_;
};
} // namespace megvii
} // namespace detection
} // namespace vision
} // namespace fastdeploy

22
csrcs/fastdeploy/vision/rangilyu/rangilyu_pybind.cc → csrcs/fastdeploy/vision/detection/contrib/yolox_pybind.cc
View File

@@ -15,27 +15,23 @@
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindRangiLyu(pybind11::module& m) {
auto rangilyu_module =
m.def_submodule("rangilyu", "https://github.com/RangiLyu/nanodet");
pybind11::class_<vision::rangilyu::NanoDetPlus, FastDeployModel>(
rangilyu_module, "NanoDetPlus")
void BindYOLOX(pybind11::module& m) {
pybind11::class_<vision::detection::YOLOX, FastDeployModel>(m, "YOLOX")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::rangilyu::NanoDetPlus& self, pybind11::array& data,
[](vision::detection::YOLOX& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::DetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::rangilyu::NanoDetPlus::size)
.def_readwrite("padding_value",
&vision::rangilyu::NanoDetPlus::padding_value)
.def_readwrite("keep_ratio", &vision::rangilyu::NanoDetPlus::keep_ratio)
.def_readwrite("size", &vision::detection::YOLOX::size)
.def_readwrite("padding_value", &vision::detection::YOLOX::padding_value)
.def_readwrite("is_decode_exported",
&vision::detection::YOLOX::is_decode_exported)
.def_readwrite("downsample_strides",
&vision::rangilyu::NanoDetPlus::downsample_strides)
.def_readwrite("max_wh", &vision::rangilyu::NanoDetPlus::max_wh)
.def_readwrite("reg_max", &vision::rangilyu::NanoDetPlus::reg_max);
&vision::detection::YOLOX::downsample_strides)
.def_readwrite("max_wh", &vision::detection::YOLOX::max_wh);
}
} // namespace fastdeploy

10
csrcs/fastdeploy/vision/detection/detection_pybind.cc
View File

@@ -19,6 +19,11 @@ namespace fastdeploy {
void BindYOLOv7(pybind11::module& m);
void BindScaledYOLOv4(pybind11::module& m);
void BindYOLOR(pybind11::module& m);
void BindYOLOv6(pybind11::module& m);
void BindYOLOv5Lite(pybind11::module& m);
void BindYOLOv5(pybind11::module& m);
void BindYOLOX(pybind11::module& m);
void BindNanoDetPlus(pybind11::module& m);
void BindPPDet(pybind11::module& m);
void BindDetection(pybind11::module& m) {
@@ -28,5 +33,10 @@ void BindDetection(pybind11::module& m) {
BindYOLOv7(detection_module);
BindScaledYOLOv4(detection_module);
BindYOLOR(detection_module);
BindYOLOv6(detection_module);
BindYOLOv5Lite(detection_module);
BindYOLOv5(detection_module);
BindYOLOX(detection_module);
BindNanoDetPlus(detection_module);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/biubug6/retinaface.cc → csrcs/fastdeploy/vision/facedet/contrib/retinaface.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/biubug6/retinaface.h"
#include "fastdeploy/vision/facedet/contrib/retinaface.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace biubug6 {
namespace facedet {
struct RetinaAnchor {
float cx;
@@ -305,6 +305,6 @@ bool RetinaFace::Predict(cv::Mat* im, FaceDetectionResult* result,
return true;
}
} // namespace biubug6
} // namespace facedet
} // namespace vision
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/biubug6/retinaface.h → csrcs/fastdeploy/vision/facedet/contrib/retinaface.h
View File

@@ -21,7 +21,7 @@ namespace fastdeploy {
namespace vision {
namespace biubug6 {
namespace facedet {
class FASTDEPLOY_DECL RetinaFace : public FastDeployModel {
public:
@@ -32,7 +32,7 @@ class FASTDEPLOY_DECL RetinaFace : public FastDeployModel {
const Frontend& model_format = Frontend::ONNX);
// 定义模型的名称
std::string ModelName() const { return "biubug6/Pytorch_Retinaface"; }
std::string ModelName() const { return "Pytorch_Retinaface"; }
// 模型预测接口，即用户调用的接口
// im 为用户的输入数据，目前对于CV均定义为cv::Mat
@@ -87,6 +87,6 @@ class FASTDEPLOY_DECL RetinaFace : public FastDeployModel {
bool is_dynamic_input_;
};
} // namespace biubug6
} // namespace facedet
} // namespace vision
} // namespace fastdeploy

18
csrcs/fastdeploy/vision/biubug6/biubug6_pybind.cc → csrcs/fastdeploy/vision/facedet/contrib/retinaface_pybind.cc
View File

@@ -15,26 +15,24 @@
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindBiubug6(pybind11::module& m) {
auto biubug6_module = m.def_submodule(
"biubug6", "https://github.com/biubug6/Pytorch_Retinaface");
pybind11::class_<vision::biubug6::RetinaFace, FastDeployModel>(biubug6_module,
void BindRetinaFace(pybind11::module& m) {
pybind11::class_<vision::facedet::RetinaFace, FastDeployModel>(m,
"RetinaFace")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::biubug6::RetinaFace& self, pybind11::array& data,
[](vision::facedet::RetinaFace& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::FaceDetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::biubug6::RetinaFace::size)
.def_readwrite("variance", &vision::biubug6::RetinaFace::variance)
.def_readwrite("size", &vision::facedet::RetinaFace::size)
.def_readwrite("variance", &vision::facedet::RetinaFace::variance)
.def_readwrite("downsample_strides",
&vision::biubug6::RetinaFace::downsample_strides)
.def_readwrite("min_sizes", &vision::biubug6::RetinaFace::min_sizes)
&vision::facedet::RetinaFace::downsample_strides)
.def_readwrite("min_sizes", &vision::facedet::RetinaFace::min_sizes)
.def_readwrite("landmarks_per_face",
&vision::biubug6::RetinaFace::landmarks_per_face);
&vision::facedet::RetinaFace::landmarks_per_face);
}
} // namespace fastdeploy

32
csrcs/fastdeploy/vision/deepinsight/scrfd.cc → csrcs/fastdeploy/vision/facedet/contrib/scrfd.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/deepinsight/scrfd.h"
#include "fastdeploy/vision/facedet/contrib/scrfd.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace facedet {
void SCRFD::LetterBox(Mat* mat, const std::vector<int>& size,
const std::vector<float>& color, bool _auto,
@@ -246,14 +246,18 @@ bool SCRFD::Postprocess(
float r = offsets[2]; // right
float b = offsets[3]; // bottom
float x1 =
((cx - l) * static_cast<float>(current_stride) - static_cast<float>(pad_w)) / scale; // cx - l x1
float y1 =
((cy - t) * static_cast<float>(current_stride) - static_cast<float>(pad_h)) / scale; // cy - t y1
float x2 =
((cx + r) * static_cast<float>(current_stride) - static_cast<float>(pad_w)) / scale; // cx + r x2
float y2 =
((cy + b) * static_cast<float>(current_stride) - static_cast<float>(pad_h)) / scale; // cy + b y2
float x1 = ((cx - l) * static_cast<float>(current_stride) -
static_cast<float>(pad_w)) /
scale; // cx - l x1
float y1 = ((cy - t) * static_cast<float>(current_stride) -
static_cast<float>(pad_h)) /
scale; // cy - t y1
float x2 = ((cx + r) * static_cast<float>(current_stride) -
static_cast<float>(pad_w)) /
scale; // cx + r x2
float y2 = ((cy + b) * static_cast<float>(current_stride) -
static_cast<float>(pad_h)) /
scale; // cy + b y2
result->boxes.emplace_back(std::array<float, 4>{x1, y1, x2, y2});
result->scores.push_back(cls_conf);
if (use_kps) {
@@ -264,9 +268,11 @@ bool SCRFD::Postprocess(
for (unsigned int j = 0; j < landmarks_per_face * 2; j += 2) {
float kps_l = kps_offsets[j];
float kps_t = kps_offsets[j + 1];
float kps_x = ((cx + kps_l) * static_cast<float>(current_stride) - static_cast<float>(pad_w)) /
float kps_x = ((cx + kps_l) * static_cast<float>(current_stride) -
static_cast<float>(pad_w)) /
scale; // cx + l x
float kps_y = ((cy + kps_t) * static_cast<float>(current_stride) - static_cast<float>(pad_h)) /
float kps_y = ((cy + kps_t) * static_cast<float>(current_stride) -
static_cast<float>(pad_h)) /
scale; // cy + t y
result->landmarks.emplace_back(std::array<float, 2>{kps_x, kps_y});
}
@@ -358,6 +364,6 @@ bool SCRFD::Predict(cv::Mat* im, FaceDetectionResult* result,
return true;
}
} // namespace deepinsight
} // namespace facedet
} // namespace vision
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/scrfd.h → csrcs/fastdeploy/vision/facedet/contrib/scrfd.h
View File

@@ -22,7 +22,7 @@ namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace facedet {
class FASTDEPLOY_DECL SCRFD : public FastDeployModel {
public:
@@ -33,7 +33,7 @@ class FASTDEPLOY_DECL SCRFD : public FastDeployModel {
const Frontend& model_format = Frontend::ONNX);
// 定义模型的名称
std::string ModelName() const { return "deepinsight/scrfd"; }
std::string ModelName() const { return "scrfd"; }
// 模型预测接口，即用户调用的接口
// im 为用户的输入数据，目前对于CV均定义为cv::Mat
@@ -117,6 +117,6 @@ class FASTDEPLOY_DECL SCRFD : public FastDeployModel {
std::unordered_map<int, std::vector<SCRFDPoint>> center_points_;
};
} // namespace deepinsight
} // namespace facedet
} // namespace vision
} // namespace fastdeploy

45
csrcs/fastdeploy/vision/facedet/contrib/scrfd_pybind.cc Normal file
View File

@@ -0,0 +1,45 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindSCRFD(pybind11::module& m) {
// Bind SCRFD
pybind11::class_<vision::facedet::SCRFD, FastDeployModel>(m, "SCRFD")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::facedet::SCRFD& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::FaceDetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::facedet::SCRFD::size)
.def_readwrite("padding_value", &vision::facedet::SCRFD::padding_value)
.def_readwrite("is_mini_pad", &vision::facedet::SCRFD::is_mini_pad)
.def_readwrite("is_no_pad", &vision::facedet::SCRFD::is_no_pad)
.def_readwrite("is_scale_up", &vision::facedet::SCRFD::is_scale_up)
.def_readwrite("stride", &vision::facedet::SCRFD::stride)
.def_readwrite("use_kps", &vision::facedet::SCRFD::use_kps)
.def_readwrite("max_nms", &vision::facedet::SCRFD::max_nms)
.def_readwrite("downsample_strides",
&vision::facedet::SCRFD::downsample_strides)
.def_readwrite("num_anchors", &vision::facedet::SCRFD::num_anchors)
.def_readwrite("landmarks_per_face",
&vision::facedet::SCRFD::landmarks_per_face);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/linzaer/ultraface.cc → csrcs/fastdeploy/vision/facedet/contrib/ultraface.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/linzaer/ultraface.h"
#include "fastdeploy/vision/facedet/contrib/ultraface.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace linzaer {
namespace facedet {
UltraFace::UltraFace(const std::string& model_file,
const std::string& params_file,
@@ -216,6 +216,6 @@ bool UltraFace::Predict(cv::Mat* im, FaceDetectionResult* result,
return true;
}
} // namespace linzaer
} // namespace facedet
} // namespace vision
} // namespace fastdeploy

4
csrcs/fastdeploy/vision/linzaer/ultraface.h → csrcs/fastdeploy/vision/facedet/contrib/ultraface.h
View File

@@ -21,7 +21,7 @@ namespace fastdeploy {
namespace vision {
namespace linzaer {
namespace facedet {
class FASTDEPLOY_DECL UltraFace : public FastDeployModel {
public:
@@ -79,6 +79,6 @@ class FASTDEPLOY_DECL UltraFace : public FastDeployModel {
bool is_dynamic_input_;
};
} // namespace linzaer
} // namespace facedet
} // namespace vision
} // namespace fastdeploy

12
csrcs/fastdeploy/vision/linzaer/linzaer_pybind.cc → csrcs/fastdeploy/vision/facedet/contrib/ultraface_pybind.cc
View File

@@ -15,21 +15,17 @@
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindLinzaer(pybind11::module& m) {
auto linzaer_module = m.def_submodule(
"linzaer",
"https://github.com/Linzaer/Ultra-Light-Fast-Generic-Face-Detector-1MB");
pybind11::class_<vision::linzaer::UltraFace, FastDeployModel>(linzaer_module,
"UltraFace")
void BindUltraFace(pybind11::module& m) {
pybind11::class_<vision::facedet::UltraFace, FastDeployModel>(m, "UltraFace")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::linzaer::UltraFace& self, pybind11::array& data,
[](vision::facedet::UltraFace& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::FaceDetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::linzaer::UltraFace::size);
.def_readwrite("size", &vision::facedet::UltraFace::size);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepcam/yolov5face.cc → csrcs/fastdeploy/vision/facedet/contrib/yolov5face.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/deepcam/yolov5face.h"
#include "fastdeploy/vision/facedet/contrib/yolov5face.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace deepcam {
namespace facedet {
void LetterBox(Mat* mat, std::vector<int> size, std::vector<float> color,
bool _auto, bool scale_fill = false, bool scale_up = true,
@@ -289,6 +289,6 @@ bool YOLOv5Face::Predict(cv::Mat* im, FaceDetectionResult* result,
return true;
}
} // namespace deepcam
} // namespace facedet
} // namespace vision
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepcam/yolov5face.h → csrcs/fastdeploy/vision/facedet/contrib/yolov5face.h
View File

@@ -21,7 +21,7 @@ namespace fastdeploy {
namespace vision {
namespace deepcam {
namespace facedet {
class FASTDEPLOY_DECL YOLOv5Face : public FastDeployModel {
public:
@@ -32,7 +32,7 @@ class FASTDEPLOY_DECL YOLOv5Face : public FastDeployModel {
const Frontend& model_format = Frontend::ONNX);
// 定义模型的名称
std::string ModelName() const { return "deepcam-cn/yolov5-face"; }
std::string ModelName() const { return "yolov5-face"; }
// 模型预测接口，即用户调用的接口
// im 为用户的输入数据，目前对于CV均定义为cv::Mat
@@ -92,6 +92,6 @@ class FASTDEPLOY_DECL YOLOv5Face : public FastDeployModel {
bool is_dynamic_input_;
};
} // namespace deepcam
} // namespace facedet
} // namespace vision
} // namespace fastdeploy

22
csrcs/fastdeploy/vision/deepcam/deepcam_pybind.cc → csrcs/fastdeploy/vision/facedet/contrib/yolov5face_pybind.cc
View File

@@ -15,29 +15,27 @@
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindDeepCam(pybind11::module& m) {
auto deepcam_module =
m.def_submodule("deepcam", "https://github.com/deepcam-cn/yolov5-face");
pybind11::class_<vision::deepcam::YOLOv5Face, FastDeployModel>(deepcam_module,
void BindYOLOv5Face(pybind11::module& m) {
pybind11::class_<vision::facedet::YOLOv5Face, FastDeployModel>(m,
"YOLOv5Face")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::deepcam::YOLOv5Face& self, pybind11::array& data,
[](vision::facedet::YOLOv5Face& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::FaceDetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::deepcam::YOLOv5Face::size)
.def_readwrite("size", &vision::facedet::YOLOv5Face::size)
.def_readwrite("padding_value",
&vision::deepcam::YOLOv5Face::padding_value)
.def_readwrite("is_mini_pad", &vision::deepcam::YOLOv5Face::is_mini_pad)
.def_readwrite("is_no_pad", &vision::deepcam::YOLOv5Face::is_no_pad)
.def_readwrite("is_scale_up", &vision::deepcam::YOLOv5Face::is_scale_up)
.def_readwrite("stride", &vision::deepcam::YOLOv5Face::stride)
&vision::facedet::YOLOv5Face::padding_value)
.def_readwrite("is_mini_pad", &vision::facedet::YOLOv5Face::is_mini_pad)
.def_readwrite("is_no_pad", &vision::facedet::YOLOv5Face::is_no_pad)
.def_readwrite("is_scale_up", &vision::facedet::YOLOv5Face::is_scale_up)
.def_readwrite("stride", &vision::facedet::YOLOv5Face::stride)
.def_readwrite("landmarks_per_face",
&vision::deepcam::YOLOv5Face::landmarks_per_face);
&vision::facedet::YOLOv5Face::landmarks_per_face);
}
} // namespace fastdeploy

31
csrcs/fastdeploy/vision/facedet/facedet_pybind.cc Normal file
View File

@@ -0,0 +1,31 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindRetinaFace(pybind11::module& m);
void BindUltraFace(pybind11::module& m);
void BindYOLOv5Face(pybind11::module& m);
void BindSCRFD(pybind11::module& m);
void BindFaceDet(pybind11::module& m) {
auto facedet_module = m.def_submodule("facedet", "Face detection models.");
BindRetinaFace(facedet_module);
BindUltraFace(facedet_module);
BindYOLOv5Face(facedet_module);
BindSCRFD(facedet_module);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/arcface.cc → csrcs/fastdeploy/vision/faceid/contrib/arcface.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/deepinsight/arcface.h"
#include "fastdeploy/vision/faceid/contrib/arcface.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
ArcFace::ArcFace(const std::string& model_file, const std::string& params_file,
const RuntimeOption& custom_option,
@@ -78,6 +78,6 @@ bool ArcFace::Predict(cv::Mat* im, FaceRecognitionResult* result) {
return InsightFaceRecognitionModel::Predict(im, result);
}
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/arcface.h → csrcs/fastdeploy/vision/faceid/contrib/arcface.h
View File

@@ -16,13 +16,13 @@
#include "fastdeploy/fastdeploy_model.h"
#include "fastdeploy/vision/common/processors/transform.h"
#include "fastdeploy/vision/common/result.h"
#include "fastdeploy/vision/deepinsight/insightface_rec.h"
#include "fastdeploy/vision/faceid/contrib/insightface_rec.h"
namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
class FASTDEPLOY_DECL ArcFace : public InsightFaceRecognitionModel {
public:
@@ -60,6 +60,6 @@ class FASTDEPLOY_DECL ArcFace : public InsightFaceRecognitionModel {
FaceRecognitionResult* result) override;
};
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

37
csrcs/fastdeploy/vision/faceid/contrib/arcface_pybind.cc Normal file
View File

@@ -0,0 +1,37 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindArcFace(pybind11::module& m) {
// Bind ArcFace
pybind11::class_<vision::faceid::ArcFace,
vision::faceid::InsightFaceRecognitionModel>(m, "ArcFace")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::faceid::ArcFace& self, pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::faceid::ArcFace::size)
.def_readwrite("alpha", &vision::faceid::ArcFace::alpha)
.def_readwrite("beta", &vision::faceid::ArcFace::beta)
.def_readwrite("swap_rb", &vision::faceid::ArcFace::swap_rb)
.def_readwrite("l2_normalize", &vision::faceid::ArcFace::l2_normalize);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/cosface.cc → csrcs/fastdeploy/vision/faceid/contrib/cosface.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/deepinsight/cosface.h"
#include "fastdeploy/vision/faceid/contrib/cosface.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
CosFace::CosFace(const std::string& model_file, const std::string& params_file,
const RuntimeOption& custom_option,
@@ -78,6 +78,6 @@ bool CosFace::Predict(cv::Mat* im, FaceRecognitionResult* result) {
return InsightFaceRecognitionModel::Predict(im, result);
}
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/cosface.h → csrcs/fastdeploy/vision/faceid/contrib/cosface.h
View File

@@ -16,13 +16,13 @@
#include "fastdeploy/fastdeploy_model.h"
#include "fastdeploy/vision/common/processors/transform.h"
#include "fastdeploy/vision/common/result.h"
#include "fastdeploy/vision/deepinsight/insightface_rec.h"
#include "fastdeploy/vision/faceid/contrib/insightface_rec.h"
namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
class FASTDEPLOY_DECL CosFace : public InsightFaceRecognitionModel {
public:
@@ -61,6 +61,6 @@ class FASTDEPLOY_DECL CosFace : public InsightFaceRecognitionModel {
FaceRecognitionResult* result) override;
};
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

37
csrcs/fastdeploy/vision/faceid/contrib/cosface_pybind.cc Normal file
View File

@@ -0,0 +1,37 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindCosFace(pybind11::module& m) {
// Bind CosFace
pybind11::class_<vision::faceid::CosFace,
vision::faceid::InsightFaceRecognitionModel>(m, "CosFace")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::faceid::CosFace& self, pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::faceid::CosFace::size)
.def_readwrite("alpha", &vision::faceid::CosFace::alpha)
.def_readwrite("beta", &vision::faceid::CosFace::beta)
.def_readwrite("swap_rb", &vision::faceid::CosFace::swap_rb)
.def_readwrite("l2_normalize", &vision::faceid::CosFace::l2_normalize);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/insightface_rec.cc → csrcs/fastdeploy/vision/faceid/contrib/insightface_rec.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/deepinsight/insightface_rec.h"
#include "fastdeploy/vision/faceid/contrib/insightface_rec.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
InsightFaceRecognitionModel::InsightFaceRecognitionModel(
const std::string& model_file, const std::string& params_file,
@@ -148,6 +148,6 @@ bool InsightFaceRecognitionModel::Predict(cv::Mat* im,
return true;
}
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

4
csrcs/fastdeploy/vision/deepinsight/insightface_rec.h → csrcs/fastdeploy/vision/faceid/contrib/insightface_rec.h
View File

@@ -21,7 +21,7 @@ namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
class FASTDEPLOY_DECL InsightFaceRecognitionModel : public FastDeployModel {
public:
@@ -67,6 +67,6 @@ class FASTDEPLOY_DECL InsightFaceRecognitionModel : public FastDeployModel {
FaceRecognitionResult* result);
};
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

42
csrcs/fastdeploy/vision/faceid/contrib/insightface_rec_pybind.cc Normal file
View File

@@ -0,0 +1,42 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindInsightFaceRecognitionModel(pybind11::module& m) {
// Bind InsightFaceRecognitionModel
pybind11::class_<vision::faceid::InsightFaceRecognitionModel,
FastDeployModel>(m, "InsightFaceRecognitionModel")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::faceid::InsightFaceRecognitionModel& self,
pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::faceid::InsightFaceRecognitionModel::size)
.def_readwrite("alpha",
&vision::faceid::InsightFaceRecognitionModel::alpha)
.def_readwrite("beta", &vision::faceid::InsightFaceRecognitionModel::beta)
.def_readwrite("swap_rb",
&vision::faceid::InsightFaceRecognitionModel::swap_rb)
.def_readwrite(
"l2_normalize",
&vision::faceid::InsightFaceRecognitionModel::l2_normalize);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/partial_fc.cc → csrcs/fastdeploy/vision/faceid/contrib/partial_fc.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/deepinsight/partial_fc.h"
#include "fastdeploy/vision/faceid/contrib/partial_fc.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
PartialFC::PartialFC(const std::string& model_file,
const std::string& params_file,
@@ -79,6 +79,6 @@ bool PartialFC::Predict(cv::Mat* im, FaceRecognitionResult* result) {
return InsightFaceRecognitionModel::Predict(im, result);
}
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/partial_fc.h → csrcs/fastdeploy/vision/faceid/contrib/partial_fc.h
View File

@@ -16,13 +16,13 @@
#include "fastdeploy/fastdeploy_model.h"
#include "fastdeploy/vision/common/processors/transform.h"
#include "fastdeploy/vision/common/result.h"
#include "fastdeploy/vision/deepinsight/insightface_rec.h"
#include "fastdeploy/vision/faceid/contrib/insightface_rec.h"
namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
class FASTDEPLOY_DECL PartialFC : public InsightFaceRecognitionModel {
public:
@@ -59,6 +59,6 @@ class FASTDEPLOY_DECL PartialFC : public InsightFaceRecognitionModel {
FaceRecognitionResult* result) override;
};
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

37
csrcs/fastdeploy/vision/faceid/contrib/partial_fc_pybind.cc Normal file
View File

@@ -0,0 +1,37 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindPartialFC(pybind11::module& m) {
// Bind Partial FC
pybind11::class_<vision::faceid::PartialFC,
vision::faceid::InsightFaceRecognitionModel>(m, "PartialFC")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::faceid::PartialFC& self, pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::faceid::PartialFC::size)
.def_readwrite("alpha", &vision::faceid::PartialFC::alpha)
.def_readwrite("beta", &vision::faceid::PartialFC::beta)
.def_readwrite("swap_rb", &vision::faceid::PartialFC::swap_rb)
.def_readwrite("l2_normalize", &vision::faceid::PartialFC::l2_normalize);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/vpl.cc → csrcs/fastdeploy/vision/faceid/contrib/vpl.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/deepinsight/vpl.h"
#include "fastdeploy/vision/faceid/contrib/vpl.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
VPL::VPL(const std::string& model_file, const std::string& params_file,
const RuntimeOption& custom_option, const Frontend& model_format)
@@ -77,6 +77,6 @@ bool VPL::Predict(cv::Mat* im, FaceRecognitionResult* result) {
return InsightFaceRecognitionModel::Predict(im, result);
}
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/deepinsight/vpl.h → csrcs/fastdeploy/vision/faceid/contrib/vpl.h
View File

@@ -16,13 +16,13 @@
#include "fastdeploy/fastdeploy_model.h"
#include "fastdeploy/vision/common/processors/transform.h"
#include "fastdeploy/vision/common/result.h"
#include "fastdeploy/vision/deepinsight/insightface_rec.h"
#include "fastdeploy/vision/faceid/contrib/insightface_rec.h"
namespace fastdeploy {
namespace vision {
namespace deepinsight {
namespace faceid {
class FASTDEPLOY_DECL VPL : public InsightFaceRecognitionModel {
public:
@@ -60,6 +60,6 @@ class FASTDEPLOY_DECL VPL : public InsightFaceRecognitionModel {
FaceRecognitionResult* result) override;
};
} // namespace deepinsight
} // namespace faceid
} // namespace vision
} // namespace fastdeploy

37
csrcs/fastdeploy/vision/faceid/contrib/vpl_pybind.cc Normal file
View File

@@ -0,0 +1,37 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindVPL(pybind11::module& m) {
// Bind VPL
pybind11::class_<vision::faceid::VPL,
vision::faceid::InsightFaceRecognitionModel>(m, "VPL")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::faceid::VPL& self, pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::FaceRecognitionResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::faceid::VPL::size)
.def_readwrite("alpha", &vision::faceid::VPL::alpha)
.def_readwrite("beta", &vision::faceid::VPL::beta)
.def_readwrite("swap_rb", &vision::faceid::VPL::swap_rb)
.def_readwrite("l2_normalize", &vision::faceid::VPL::l2_normalize);
}
} // namespace fastdeploy

33
csrcs/fastdeploy/vision/faceid/faceid_pybind.cc Normal file
View File

@@ -0,0 +1,33 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindArcFace(pybind11::module& m);
void BindInsightFaceRecognitionModel(pybind11::module& m);
void BindCosFace(pybind11::module& m);
void BindPartialFC(pybind11::module& m);
void BindVPL(pybind11::module& m);
void BindFaceId(pybind11::module& m) {
auto faceid_module = m.def_submodule("faceid", "Face recognition models.");
BindInsightFaceRecognitionModel(faceid_module);
BindArcFace(faceid_module);
BindCosFace(faceid_module);
BindPartialFC(faceid_module);
BindVPL(faceid_module);
}
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/zhkkke/modnet.cc → csrcs/fastdeploy/vision/matting/contrib/modnet.cc
View File

@@ -12,7 +12,7 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/vision/zhkkke/modnet.h"
#include "fastdeploy/vision/matting/contrib/modnet.h"
#include "fastdeploy/utils/perf.h"
#include "fastdeploy/vision/utils/utils.h"
@@ -20,7 +20,7 @@ namespace fastdeploy {
namespace vision {
namespace zhkkke {
namespace matting {
MODNet::MODNet(const std::string& model_file, const std::string& params_file,
const RuntimeOption& custom_option,
@@ -170,6 +170,6 @@ bool MODNet::Predict(cv::Mat* im, MattingResult* result) {
return true;
}
} // namespace zhkkke
} // namespace matting
} // namespace vision
} // namespace fastdeploy

6
csrcs/fastdeploy/vision/zhkkke/modnet.h → csrcs/fastdeploy/vision/matting/contrib/modnet.h
View File

@@ -21,7 +21,7 @@ namespace fastdeploy {
namespace vision {
namespace zhkkke {
namespace matting {
class FASTDEPLOY_DECL MODNet : public FastDeployModel {
public:
@@ -32,7 +32,7 @@ class FASTDEPLOY_DECL MODNet : public FastDeployModel {
const Frontend& model_format = Frontend::ONNX);
// 定义模型的名称
std::string ModelName() const { return "ZHKKKe/MODNet"; }
std::string ModelName() const { return "matting/MODNet"; }
// 以下为一些可供用户修改的属性
// tuple of (width, height), default (256, 256)
@@ -65,6 +65,6 @@ class FASTDEPLOY_DECL MODNet : public FastDeployModel {
const std::map<std::string, std::array<int, 2>>& im_info);
};
} // namespace zhkkke
} // namespace matting
} // namespace vision
} // namespace fastdeploy

16
csrcs/fastdeploy/vision/zhkkke/zhkkke_pybind.cc → csrcs/fastdeploy/vision/matting/contrib/modnet_pybind.cc
View File

@@ -15,23 +15,21 @@
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindZHKKKe(pybind11::module& m) {
auto zhkkke_module = m.def_submodule("zhkkke", "https://github.com/ZHKKKe");
void BindMODNet(pybind11::module& m) {
// Bind MODNet
pybind11::class_<vision::zhkkke::MODNet, FastDeployModel>(zhkkke_module,
"MODNet")
pybind11::class_<vision::matting::MODNet, FastDeployModel>(m, "MODNet")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::zhkkke::MODNet& self, pybind11::array& data) {
[](vision::matting::MODNet& self, pybind11::array& data) {
auto mat = PyArrayToCvMat(data);
vision::MattingResult res;
self.Predict(&mat, &res);
return res;
})
.def_readwrite("size", &vision::zhkkke::MODNet::size)
.def_readwrite("alpha", &vision::zhkkke::MODNet::alpha)
.def_readwrite("beta", &vision::zhkkke::MODNet::beta)
.def_readwrite("swap_rb", &vision::zhkkke::MODNet::swap_rb);
.def_readwrite("size", &vision::matting::MODNet::size)
.def_readwrite("alpha", &vision::matting::MODNet::alpha)
.def_readwrite("beta", &vision::matting::MODNet::beta)
.def_readwrite("swap_rb", &vision::matting::MODNet::swap_rb);
}
} // namespace fastdeploy

26
csrcs/fastdeploy/vision/matting/matting_pybind.cc Normal file
View File

@@ -0,0 +1,26 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindMODNet(pybind11::module& m);
void BindMatting(pybind11::module& m) {
auto matting_module =
m.def_submodule("matting", "Image object matting models.");
BindMODNet(matting_module);
}
} // namespace fastdeploy

42
csrcs/fastdeploy/vision/ultralytics/ultralytics_pybind.cc
View File

@@ -1,42 +0,0 @@
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "fastdeploy/pybind/main.h"
namespace fastdeploy {
void BindUltralytics(pybind11::module& m) {
auto ultralytics_module =
m.def_submodule("ultralytics", "https://github.com/ultralytics/yolov5");
pybind11::class_<vision::ultralytics::YOLOv5, FastDeployModel>(
ultralytics_module, "YOLOv5")
.def(pybind11::init<std::string, std::string, RuntimeOption, Frontend>())
.def("predict",
[](vision::ultralytics::YOLOv5& self, pybind11::array& data,
float conf_threshold, float nms_iou_threshold) {
auto mat = PyArrayToCvMat(data);
vision::DetectionResult res;
self.Predict(&mat, &res, conf_threshold, nms_iou_threshold);
return res;
})
.def_readwrite("size", &vision::ultralytics::YOLOv5::size)
.def_readwrite("padding_value",
&vision::ultralytics::YOLOv5::padding_value)
.def_readwrite("is_mini_pad", &vision::ultralytics::YOLOv5::is_mini_pad)
.def_readwrite("is_no_pad", &vision::ultralytics::YOLOv5::is_no_pad)
.def_readwrite("is_scale_up", &vision::ultralytics::YOLOv5::is_scale_up)
.def_readwrite("stride", &vision::ultralytics::YOLOv5::stride)
.def_readwrite("max_wh", &vision::ultralytics::YOLOv5::max_wh)
.def_readwrite("multi_label", &vision::ultralytics::YOLOv5::multi_label);
}
} // namespace fastdeploy

26
csrcs/fastdeploy/vision/vision_pybind.cc
View File

@@ -19,18 +19,11 @@ namespace fastdeploy {
void BindPPCls(pybind11::module& m);
void BindPPDet(pybind11::module& m);
void BindPPSeg(pybind11::module& m);
void BindUltralytics(pybind11::module& m);
void BindMeituan(pybind11::module& m);
void BindMegvii(pybind11::module& m);
void BindDeepCam(pybind11::module& m);
void BindRangiLyu(pybind11::module& m);
void BindLinzaer(pybind11::module& m);
void BindBiubug6(pybind11::module& m);
void BindPpogg(pybind11::module& m);
void BindDeepInsight(pybind11::module& m);
void BindZHKKKe(pybind11::module& m);
void BindDetection(pybind11::module& m);
void BindMatting(pybind11::module& m);
void BindFaceDet(pybind11::module& m);
void BindFaceId(pybind11::module& m);
#ifdef ENABLE_VISION_VISUALIZE
void BindVisualize(pybind11::module& m);
#endif
@@ -88,18 +81,11 @@ void BindVision(pybind11::module& m) {
BindPPCls(m);
BindPPDet(m);
BindPPSeg(m);
BindUltralytics(m);
BindMeituan(m);
BindMegvii(m);
BindDeepCam(m);
BindRangiLyu(m);
BindLinzaer(m);
BindBiubug6(m);
BindPpogg(m);
BindDeepInsight(m);
BindZHKKKe(m);
BindDetection(m);
BindFaceDet(m);
BindFaceId(m);
BindMatting(m);
#ifdef ENABLE_VISION_VISUALIZE
BindVisualize(m);
#endif

4
csrcs/fastdeploy/vision/visualize/visualize.h
View File

@@ -26,10 +26,10 @@ class FASTDEPLOY_DECL Visualize {
static std::vector<int> color_map_;
static const std::vector<int>& GetColorMap(int num_classes = 1000);
static cv::Mat VisDetection(const cv::Mat& im, const DetectionResult& result,
int line_size = 2, float font_size = 0.5f);
int line_size = 1, float font_size = 0.5f);
static cv::Mat VisFaceDetection(const cv::Mat& im,
const FaceDetectionResult& result,
int line_size = 2, float font_size = 0.5f);
int line_size = 1, float font_size = 0.5f);
static cv::Mat VisSegmentation(const cv::Mat& im,
const SegmentationResult& result);
static cv::Mat VisMattingAlpha(const cv::Mat& im, const MattingResult& result,

18
fastdeploy/vision/__init__.py
View File

@@ -13,21 +13,9 @@
# limitations under the License.
from __future__ import absolute_import
from . import evaluation
from . import ppcls
from . import ppdet
from . import ppseg
from . import ultralytics
from . import meituan
from . import megvii
from . import deepcam
from . import rangilyu
from . import linzaer
from . import biubug6
from . import ppogg
from . import deepinsight
from . import zhkkke
from . import detection
from . import matting
from . import facedet
from . import faceid
from .visualize import *

563
fastdeploy/vision/deepinsight/__init__.py
View File

@@ -1,563 +0,0 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
class SCRFD(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(SCRFD, self).__init__(runtime_option)
self._model = C.vision.deepinsight.SCRFD(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "SCRFD initialize failed."
def predict(self, input_image, conf_threshold=0.7, nms_iou_threshold=0.3):
return self._model.predict(input_image, conf_threshold,
nms_iou_threshold)
# 一些跟SCRFD模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [640, 640]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def padding_value(self):
return self._model.padding_value
@property
def is_no_pad(self):
return self._model.is_no_pad
@property
def is_mini_pad(self):
return self._model.is_mini_pad
@property
def is_scale_up(self):
return self._model.is_scale_up
@property
def stride(self):
return self._model.stride
@property
def downsample_strides(self):
return self._model.downsample_strides
@property
def landmarks_per_face(self):
return self._model.landmarks_per_face
@property
def use_kps(self):
return self._model.use_kps
@property
def max_nms(self):
return self._model.max_nms
@property
def num_anchors(self):
return self._model.num_anchors
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@padding_value.setter
def padding_value(self, value):
assert isinstance(
value,
list), "The value to set `padding_value` must be type of list."
self._model.padding_value = value
@is_no_pad.setter
def is_no_pad(self, value):
assert isinstance(
value, bool), "The value to set `is_no_pad` must be type of bool."
self._model.is_no_pad = value
@is_mini_pad.setter
def is_mini_pad(self, value):
assert isinstance(
value,
bool), "The value to set `is_mini_pad` must be type of bool."
self._model.is_mini_pad = value
@is_scale_up.setter
def is_scale_up(self, value):
assert isinstance(
value,
bool), "The value to set `is_scale_up` must be type of bool."
self._model.is_scale_up = value
@stride.setter
def stride(self, value):
assert isinstance(
value, int), "The value to set `stride` must be type of int."
self._model.stride = value
@downsample_strides.setter
def downsample_strides(self, value):
assert isinstance(
value,
list), "The value to set `downsample_strides` must be type of list."
self._model.downsample_strides = value
@landmarks_per_face.setter
def landmarks_per_face(self, value):
assert isinstance(
value,
int), "The value to set `landmarks_per_face` must be type of int."
self._model.landmarks_per_face = value
@use_kps.setter
def use_kps(self, value):
assert isinstance(
value, bool), "The value to set `use_kps` must be type of bool."
self._model.use_kps = value
@max_nms.setter
def max_nms(self, value):
assert isinstance(
value, int), "The value to set `max_nms` must be type of int."
self._model.max_nms = value
@num_anchors.setter
def num_anchors(self, value):
assert isinstance(
value, int), "The value to set `num_anchors` must be type of int."
self._model.num_anchors = value
class InsightFaceRecognitionModel(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(InsightFaceRecognitionModel, self).__init__(runtime_option)
self._model = C.vision.deepinsight.InsightFaceRecognitionModel(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "InsightFaceRecognitionModel initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟InsightFaceRecognitionModel模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value
class ArcFace(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(ArcFace, self).__init__(runtime_option)
self._model = C.vision.deepinsight.ArcFace(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "ArcFace initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value
class CosFace(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(CosFace, self).__init__(runtime_option)
self._model = C.vision.deepinsight.CosFace(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "CosFace initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value
class PartialFC(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(PartialFC, self).__init__(runtime_option)
self._model = C.vision.deepinsight.PartialFC(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "PartialFC initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value
class VPL(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(VPL, self).__init__(runtime_option)
self._model = C.vision.deepinsight.VPL(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "VPL initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value

12
fastdeploy/vision/detection/__init__.py
View File

@@ -13,8 +13,12 @@
# limitations under the License.
from __future__ import absolute_import
from .yolov7 import YOLOv7
from .yolor import YOLOR
from .scaled_yolov4 import ScaledYOLOv4
from .contrib.yolov7 import YOLOv7
from .contrib.yolor import YOLOR
from .contrib.scaled_yolov4 import ScaledYOLOv4
from .contrib.nanodet_plus import NanoDetPlus
from .contrib.yolox import YOLOX
from .contrib.yolov5 import YOLOv5
from .contrib.yolov5lite import YOLOv5Lite
from .contrib.yolov6 import YOLOv6
from .ppdet import PPYOLOE, PPYOLO, PPYOLOv2, PaddleYOLOX, PicoDet, FasterRCNN, YOLOv3

15
fastdeploy/vision/detection/contrib/__init__.py Normal file
View File

@@ -0,0 +1,15 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import

6
fastdeploy/vision/rangilyu/__init__.py → fastdeploy/vision/detection/contrib/nanodet_plus.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class NanoDetPlus(FastDeployModel):
@@ -28,7 +28,7 @@ class NanoDetPlus(FastDeployModel):
# 初始化后的option保存在self._runtime_option
super(NanoDetPlus, self).__init__(runtime_option)
self._model = C.vision.rangilyu.NanoDetPlus(
self._model = C.vision.detection.NanoDetPlus(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "NanoDetPlus initialize failed."

4
fastdeploy/vision/detection/scaled_yolov4.py → fastdeploy/vision/detection/contrib/scaled_yolov4.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class ScaledYOLOv4(FastDeployModel):

4
fastdeploy/vision/detection/yolor.py → fastdeploy/vision/detection/contrib/yolor.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class YOLOR(FastDeployModel):

6
fastdeploy/vision/ultralytics/__init__.py → fastdeploy/vision/detection/contrib/yolov5.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class YOLOv5(FastDeployModel):
@@ -28,7 +28,7 @@ class YOLOv5(FastDeployModel):
# 初始化后的option保存在self._runtime_option
super(YOLOv5, self).__init__(runtime_option)
self._model = C.vision.ultralytics.YOLOv5(
self._model = C.vision.detection.YOLOv5(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "YOLOv5 initialize failed."

6
fastdeploy/vision/ppogg/__init__.py → fastdeploy/vision/detection/contrib/yolov5lite.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class YOLOv5Lite(FastDeployModel):
@@ -28,7 +28,7 @@ class YOLOv5Lite(FastDeployModel):
# 初始化后的option保存在self._runtime_option
super(YOLOv5Lite, self).__init__(runtime_option)
self._model = C.vision.ppogg.YOLOv5Lite(
self._model = C.vision.detection.YOLOv5Lite(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "YOLOv5Lite initialize failed."

6
fastdeploy/vision/meituan/__init__.py → fastdeploy/vision/detection/contrib/yolov6.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class YOLOv6(FastDeployModel):
@@ -28,7 +28,7 @@ class YOLOv6(FastDeployModel):
# 初始化后的option保存在self._runtime_option
super(YOLOv6, self).__init__(runtime_option)
self._model = C.vision.meituan.YOLOv6(
self._model = C.vision.detection.YOLOv6(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "YOLOv6 initialize failed."

4
fastdeploy/vision/detection/yolov7.py → fastdeploy/vision/detection/contrib/yolov7.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class YOLOv7(FastDeployModel):

8
fastdeploy/vision/megvii/__init__.py → fastdeploy/vision/detection/contrib/yolox.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class YOLOX(FastDeployModel):
@@ -28,8 +28,8 @@ class YOLOX(FastDeployModel):
# 初始化后的option保存在self._runtime_option
super(YOLOX, self).__init__(runtime_option)
self._model = C.vision.megvii.YOLOX(model_file, params_file,
self._runtime_option, model_format)
self._model = C.vision.detection.YOLOX(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "YOLOX initialize failed."

19
fastdeploy/vision/facedet/__init__.py Normal file
View File

@@ -0,0 +1,19 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from .contrib.yolov5face import YOLOv5Face
from .contrib.retinaface import RetinaFace
from .contrib.scrfd import SCRFD
from .contrib.ultraface import UltraFace

15
fastdeploy/vision/facedet/contrib/__init__.py Normal file
View File

@@ -0,0 +1,15 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import

6
fastdeploy/vision/biubug6/__init__.py → fastdeploy/vision/facedet/contrib/retinaface.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class RetinaFace(FastDeployModel):
@@ -28,7 +28,7 @@ class RetinaFace(FastDeployModel):
# 初始化后的option保存在self._runtime_option
super(RetinaFace, self).__init__(runtime_option)
self._model = C.vision.biubug6.RetinaFace(
self._model = C.vision.facedet.RetinaFace(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "RetinaFace initialize failed."

158
fastdeploy/vision/facedet/contrib/scrfd.py Normal file
View File

@@ -0,0 +1,158 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
import logging
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class SCRFD(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(SCRFD, self).__init__(runtime_option)
self._model = C.vision.facedet.SCRFD(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "SCRFD initialize failed."
def predict(self, input_image, conf_threshold=0.7, nms_iou_threshold=0.3):
return self._model.predict(input_image, conf_threshold,
nms_iou_threshold)
# 一些跟SCRFD模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [640, 640]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def padding_value(self):
return self._model.padding_value
@property
def is_no_pad(self):
return self._model.is_no_pad
@property
def is_mini_pad(self):
return self._model.is_mini_pad
@property
def is_scale_up(self):
return self._model.is_scale_up
@property
def stride(self):
return self._model.stride
@property
def downsample_strides(self):
return self._model.downsample_strides
@property
def landmarks_per_face(self):
return self._model.landmarks_per_face
@property
def use_kps(self):
return self._model.use_kps
@property
def max_nms(self):
return self._model.max_nms
@property
def num_anchors(self):
return self._model.num_anchors
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@padding_value.setter
def padding_value(self, value):
assert isinstance(
value,
list), "The value to set `padding_value` must be type of list."
self._model.padding_value = value
@is_no_pad.setter
def is_no_pad(self, value):
assert isinstance(
value, bool), "The value to set `is_no_pad` must be type of bool."
self._model.is_no_pad = value
@is_mini_pad.setter
def is_mini_pad(self, value):
assert isinstance(
value,
bool), "The value to set `is_mini_pad` must be type of bool."
self._model.is_mini_pad = value
@is_scale_up.setter
def is_scale_up(self, value):
assert isinstance(
value,
bool), "The value to set `is_scale_up` must be type of bool."
self._model.is_scale_up = value
@stride.setter
def stride(self, value):
assert isinstance(
value, int), "The value to set `stride` must be type of int."
self._model.stride = value
@downsample_strides.setter
def downsample_strides(self, value):
assert isinstance(
value,
list), "The value to set `downsample_strides` must be type of list."
self._model.downsample_strides = value
@landmarks_per_face.setter
def landmarks_per_face(self, value):
assert isinstance(
value,
int), "The value to set `landmarks_per_face` must be type of int."
self._model.landmarks_per_face = value
@use_kps.setter
def use_kps(self, value):
assert isinstance(
value, bool), "The value to set `use_kps` must be type of bool."
self._model.use_kps = value
@max_nms.setter
def max_nms(self, value):
assert isinstance(
value, int), "The value to set `max_nms` must be type of int."
self._model.max_nms = value
@num_anchors.setter
def num_anchors(self, value):
assert isinstance(
value, int), "The value to set `num_anchors` must be type of int."
self._model.num_anchors = value

6
fastdeploy/vision/linzaer/__init__.py → fastdeploy/vision/facedet/contrib/ultraface.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class UltraFace(FastDeployModel):
@@ -28,7 +28,7 @@ class UltraFace(FastDeployModel):
# 初始化后的option保存在self._runtime_option
super(UltraFace, self).__init__(runtime_option)
self._model = C.vision.linzaer.UltraFace(
self._model = C.vision.facedet.UltraFace(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "UltraFace initialize failed."

6
fastdeploy/vision/deepcam/__init__.py → fastdeploy/vision/facedet/contrib/yolov5face.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class YOLOv5Face(FastDeployModel):
@@ -28,7 +28,7 @@ class YOLOv5Face(FastDeployModel):
# 初始化后的option保存在self._runtime_option
super(YOLOv5Face, self).__init__(runtime_option)
self._model = C.vision.deepcam.YOLOv5Face(
self._model = C.vision.facedet.YOLOv5Face(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "YOLOv5Face initialize failed."

20
fastdeploy/vision/faceid/__init__.py Normal file
View File

@@ -0,0 +1,20 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from .contrib.arcface import ArcFace
from .contrib.cosface import CosFace
from .contrib.insightface_rec import InsightFaceRecognitionModel
from .contrib.partial_fc import PartialFC
from .contrib.vpl import VPL

15
fastdeploy/vision/faceid/contrib/__init__.py Normal file
View File

@@ -0,0 +1,15 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import

100
fastdeploy/vision/faceid/contrib/arcface.py Normal file
View File

@@ -0,0 +1,100 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
import logging
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
from ..contrib.insightface_rec import InsightFaceRecognitionModel
class ArcFace(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(ArcFace, self).__init__(runtime_option)
self._model = C.vision.faceid.ArcFace(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "ArcFace initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value

99
fastdeploy/vision/faceid/contrib/cosface.py Normal file
View File

@@ -0,0 +1,99 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
import logging
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class CosFace(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(CosFace, self).__init__(runtime_option)
self._model = C.vision.faceid.CosFace(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "CosFace initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value

99
fastdeploy/vision/faceid/contrib/insightface_rec.py Normal file
View File

@@ -0,0 +1,99 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
import logging
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class InsightFaceRecognitionModel(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(InsightFaceRecognitionModel, self).__init__(runtime_option)
self._model = C.vision.faceid.InsightFaceRecognitionModel(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "InsightFaceRecognitionModel initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟InsightFaceRecognitionModel模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value

99
fastdeploy/vision/faceid/contrib/partial_fc.py Normal file
View File

@@ -0,0 +1,99 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
import logging
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class PartialFC(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(PartialFC, self).__init__(runtime_option)
self._model = C.vision.faceid.PartialFC(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "PartialFC initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value

99
fastdeploy/vision/faceid/contrib/vpl.py Normal file
View File

@@ -0,0 +1,99 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
import logging
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class VPL(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(VPL, self).__init__(runtime_option)
self._model = C.vision.faceid.VPL(model_file, params_file,
self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "VPL initialize failed."
def predict(self, input_image):
return self._model.predict(input_image)
# 一些跟模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [112, 112]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def alpha(self):
return self._model.alpha
@property
def beta(self):
return self._model.beta
@property
def swap_rb(self):
return self._model.swap_rb
@property
def l2_normalize(self):
return self._model.l2_normalize
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@alpha.setter
def alpha(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `alpha` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `alpha` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.alpha = value
@beta.setter
def beta(self, value):
assert isinstance(value, (list, tuple)),\
"The value to set `beta` must be type of tuple or list."
assert len(value) == 3,\
"The value to set `beta` must contatins 3 elements for each channels, but now it contains {} elements.".format(
len(value))
self._model.beta = value
@swap_rb.setter
def swap_rb(self, value):
assert isinstance(
value, bool), "The value to set `swap_rb` must be type of bool."
self._model.swap_rb = value
@l2_normalize.setter
def l2_normalize(self, value):
assert isinstance(
value,
bool), "The value to set `l2_normalize` must be type of bool."
self._model.l2_normalize = value

16
fastdeploy/vision/matting/__init__.py Normal file
View File

@@ -0,0 +1,16 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from .contrib.modnet import MODNet

15
fastdeploy/vision/matting/contrib/__init__.py Normal file
View File

@@ -0,0 +1,15 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import

6
fastdeploy/vision/zhkkke/__init__.py → fastdeploy/vision/matting/contrib/modnet.py
View File

@@ -14,8 +14,8 @@
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
from .... import FastDeployModel, Frontend
from .... import c_lib_wrap as C
class MODNet(FastDeployModel):
@@ -28,7 +28,7 @@ class MODNet(FastDeployModel):
# 初始化后的option保存在self._runtime_option
super(MODNet, self).__init__(runtime_option)
self._model = C.vision.zhkkke.MODNet(
self._model = C.vision.matting.MODNet(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "MODNet initialize failed."

4
fastdeploy/vision/visualize/__init__.py
View File

@@ -17,12 +17,12 @@ import logging
from ... import c_lib_wrap as C
def vis_detection(im_data, det_result, line_size=2, font_size=0.5):
def vis_detection(im_data, det_result, line_size=1, font_size=0.5):
return C.vision.Visualize.vis_detection(im_data, det_result, line_size,
font_size)
def vis_face_detection(im_data, face_det_result, line_size=2, font_size=0.5):
def vis_face_detection(im_data, face_det_result, line_size=1, font_size=0.5):
return C.vision.Visualize.vis_face_detection(im_data, face_det_result,
line_size, font_size)

312
fastdeploy/vision/wongkinyiu/__init__.py
View File

@@ -1,312 +0,0 @@
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
import logging
from ... import FastDeployModel, Frontend
from ... import c_lib_wrap as C
class YOLOv7(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(YOLOv7, self).__init__(runtime_option)
self._model = C.vision.wongkinyiu.YOLOv7(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "YOLOv7 initialize failed."
def predict(self, input_image, conf_threshold=0.25, nms_iou_threshold=0.5):
return self._model.predict(input_image, conf_threshold,
nms_iou_threshold)
# 一些跟YOLOv7模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [1280, 1280]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def padding_value(self):
return self._model.padding_value
@property
def is_no_pad(self):
return self._model.is_no_pad
@property
def is_mini_pad(self):
return self._model.is_mini_pad
@property
def is_scale_up(self):
return self._model.is_scale_up
@property
def stride(self):
return self._model.stride
@property
def max_wh(self):
return self._model.max_wh
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@padding_value.setter
def padding_value(self, value):
assert isinstance(
value,
list), "The value to set `padding_value` must be type of list."
self._model.padding_value = value
@is_no_pad.setter
def is_no_pad(self, value):
assert isinstance(
value, bool), "The value to set `is_no_pad` must be type of bool."
self._model.is_no_pad = value
@is_mini_pad.setter
def is_mini_pad(self, value):
assert isinstance(
value,
bool), "The value to set `is_mini_pad` must be type of bool."
self._model.is_mini_pad = value
@is_scale_up.setter
def is_scale_up(self, value):
assert isinstance(
value,
bool), "The value to set `is_scale_up` must be type of bool."
self._model.is_scale_up = value
@stride.setter
def stride(self, value):
assert isinstance(
value, int), "The value to set `stride` must be type of int."
self._model.stride = value
@max_wh.setter
def max_wh(self, value):
assert isinstance(
value, float), "The value to set `max_wh` must be type of float."
self._model.max_wh = value
class YOLOR(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(YOLOR, self).__init__(runtime_option)
self._model = C.vision.wongkinyiu.YOLOR(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "YOLOR initialize failed."
def predict(self, input_image, conf_threshold=0.25, nms_iou_threshold=0.5):
return self._model.predict(input_image, conf_threshold,
nms_iou_threshold)
# 一些跟YOLOR模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [1280, 1280]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def padding_value(self):
return self._model.padding_value
@property
def is_no_pad(self):
return self._model.is_no_pad
@property
def is_mini_pad(self):
return self._model.is_mini_pad
@property
def is_scale_up(self):
return self._model.is_scale_up
@property
def stride(self):
return self._model.stride
@property
def max_wh(self):
return self._model.max_wh
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@padding_value.setter
def padding_value(self, value):
assert isinstance(
value,
list), "The value to set `padding_value` must be type of list."
self._model.padding_value = value
@is_no_pad.setter
def is_no_pad(self, value):
assert isinstance(
value, bool), "The value to set `is_no_pad` must be type of bool."
self._model.is_no_pad = value
@is_mini_pad.setter
def is_mini_pad(self, value):
assert isinstance(
value,
bool), "The value to set `is_mini_pad` must be type of bool."
self._model.is_mini_pad = value
@is_scale_up.setter
def is_scale_up(self, value):
assert isinstance(
value,
bool), "The value to set `is_scale_up` must be type of bool."
self._model.is_scale_up = value
@stride.setter
def stride(self, value):
assert isinstance(
value, int), "The value to set `stride` must be type of int."
self._model.stride = value
@max_wh.setter
def max_wh(self, value):
assert isinstance(
value, float), "The value to set `max_wh` must be type of float."
self._model.max_wh = value
class ScaledYOLOv4(FastDeployModel):
def __init__(self,
model_file,
params_file="",
runtime_option=None,
model_format=Frontend.ONNX):
# 调用基函数进行backend_option的初始化
# 初始化后的option保存在self._runtime_option
super(ScaledYOLOv4, self).__init__(runtime_option)
self._model = C.vision.wongkinyiu.ScaledYOLOv4(
model_file, params_file, self._runtime_option, model_format)
# 通过self.initialized判断整个模型的初始化是否成功
assert self.initialized, "ScaledYOLOv4 initialize failed."
def predict(self, input_image, conf_threshold=0.25, nms_iou_threshold=0.5):
return self._model.predict(input_image, conf_threshold,
nms_iou_threshold)
# 一些跟ScaledYOLOv4模型有关的属性封装
# 多数是预处理相关，可通过修改如model.size = [1280, 1280]改变预处理时resize的大小（前提是模型支持）
@property
def size(self):
return self._model.size
@property
def padding_value(self):
return self._model.padding_value
@property
def is_no_pad(self):
return self._model.is_no_pad
@property
def is_mini_pad(self):
return self._model.is_mini_pad
@property
def is_scale_up(self):
return self._model.is_scale_up
@property
def stride(self):
return self._model.stride
@property
def max_wh(self):
return self._model.max_wh
@size.setter
def size(self, wh):
assert isinstance(wh, (list, tuple)),\
"The value to set `size` must be type of tuple or list."
assert len(wh) == 2,\
"The value to set `size` must contatins 2 elements means [width, height], but now it contains {} elements.".format(
len(wh))
self._model.size = wh
@padding_value.setter
def padding_value(self, value):
assert isinstance(
value,
list), "The value to set `padding_value` must be type of list."
self._model.padding_value = value
@is_no_pad.setter
def is_no_pad(self, value):
assert isinstance(
value, bool), "The value to set `is_no_pad` must be type of bool."
self._model.is_no_pad = value
@is_mini_pad.setter
def is_mini_pad(self, value):
assert isinstance(
value,
bool), "The value to set `is_mini_pad` must be type of bool."
self._model.is_mini_pad = value
@is_scale_up.setter
def is_scale_up(self, value):
assert isinstance(
value,
bool), "The value to set `is_scale_up` must be type of bool."
self._model.is_scale_up = value
@stride.setter
def stride(self, value):
assert isinstance(
value, int), "The value to set `stride` must be type of int."
self._model.stride = value
@max_wh.setter
def max_wh(self, value):
assert isinstance(
value, float), "The value to set `max_wh` must be type of float."
self._model.max_wh = value

2
new_examples/vision/detection/yolov7/cpp/README.md
View File

@@ -35,7 +35,7 @@ wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/0000000
### YOLOv7类
```
fastdeploy::vision::wongkinyiu::YOLOv7(
fastdeploy::vision::detection::YOLOv7(
const string& model_file,
const string& params_file = "",
const RuntimeOption& runtime_option = RuntimeOption(),

13
new_examples/vision/detection/yolov7/cpp/infer.cc
View File

@@ -35,7 +35,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) {
std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl;
}
void GpuInfer() {
void GpuInfer(const std::string& model_file, const std::string& image_file) {
auto option = fastdeploy::RuntimeOption();
option.UseGpu();
auto model = fastdeploy::vision::detection::YOLOv7(model_file, "", option);
@@ -58,12 +58,11 @@ void GpuInfer() {
std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl;
}
void TrtInfer() {
void TrtInfer(const std::string& model_file, const std::string& image_file) {
auto option = fastdeploy::RuntimeOption();
option.UseGpu();
option.UseTrtBackend();
option.SetTrtInputShape("image", {1, 3, 320, 320}, {1, 3, 640, 640},
{1, 3, 1280, 1280});
option.SetTrtInputShape("images", {1, 3, 640, 640});
auto model = fastdeploy::vision::detection::YOLOv7(model_file, "", option);
if (!model.Initialized()) {
std::cerr << "Failed to initialize." << std::endl;
@@ -96,11 +95,11 @@ int main(int argc, char* argv[]) {
}
if (std::atoi(argv[3]) == 0) {
CpuInfer();
CpuInfer(argv[1], argv[2]);
} else if (std::atoi(argv[3]) == 1) {
GpuInfer();
GpuInfer(argv[1], argv[2]);
} else if (std::atoi(argv[3]) == 2) {
TrtInfer();
TrtInfer(argv[1], argv[2]);
}
return 0;
}

3
new_examples/vision/detection/yolov7/python/README.md
View File

@@ -10,7 +10,8 @@
```
#下载yolov7模型文件和测试图片
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov7.onnx
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000087038.jpg
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
#下载部署示例代码
git clone https://github.com/PaddlePaddle/FastDeploy.git

3
new_examples/vision/detection/yolov7/python/infer.py
View File

@@ -31,8 +31,7 @@ def build_option(args):
if args.use_trt:
option.use_trt_backend()
option.set_trt_input_shape("image", [1, 3, 320, 320],
[1, 3, 640, 640], [1, 3, 1280, 1280])
option.set_trt_input_shape("images", [1, 3, 640, 640])
return option