[Model] Add YOLOv5-seg (#988)

* add onnx_ort_runtime demo

* rm in requirements

* support batch eval

* fixed MattingResults bug

* move assignment for DetectionResult

* integrated x2paddle

* add model convert readme

* update readme

* re-lint

* add processor api

* Add MattingResult Free

* change valid_cpu_backends order

* add ppocr benchmark

* mv bs from 64 to 32

* fixed quantize.md

* fixed quantize bugs

* Add Monitor for benchmark

* update mem monitor

* Set trt_max_batch_size default 1

* fixed ocr benchmark bug

* support yolov5 in serving

* Fixed yolov5 serving

* Fixed postprocess

* update yolov5 to 7.0

* add poros runtime demos

* update readme

* Support poros abi=1

* rm useless note

* deal with comments

* support pp_trt for ppseg

* fixed symlink problem

* Add is_mini_pad and stride for yolov5

* Add yolo series for paddle format

* fixed bugs

* fixed bug

* support yolov5seg

* fixed bug

* refactor yolov5seg

* fixed bug

* mv Mask int32 to uint8

* add yolov5seg example

* rm log info

* fixed code style

* add yolov5seg example in python

* fixed dtype bug

* update note

* deal with comments

* get sorted index

* add yolov5seg test case

* Add GPL-3.0 License

* add round func

* deal with comments

* deal with commens

Co-authored-by: Jason <jiangjiajun@baidu.com>

fastdeploy/vision/detection/contrib/yolov5seg/preprocessor.cc

@@ -0,0 +1,116 @@
+// 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/vision/detection/contrib/yolov5seg/preprocessor.h"
+#include "fastdeploy/function/concat.h"
+
+namespace fastdeploy {
+namespace vision {
+namespace detection {
+
+YOLOv5SegPreprocessor::YOLOv5SegPreprocessor() {
+  size_ = {640, 640};
+  padding_value_ = {114.0, 114.0, 114.0};
+  is_mini_pad_ = false;
+  is_no_pad_ = false;
+  is_scale_up_ = true;
+  stride_ = 32;
+  max_wh_ = 7680.0;
+}
+
+void YOLOv5SegPreprocessor::LetterBox(FDMat* mat) {
+  float scale =
+      std::min(size_[1] * 1.0 / mat->Height(), size_[0] * 1.0 / mat->Width());
+  if (!is_scale_up_) {
+    scale = std::min(scale, 1.0f);
+  }
+
+  int resize_h = int(round(mat->Height() * scale));
+  int resize_w = int(round(mat->Width() * scale));
+
+  int pad_w = size_[0] - resize_w;
+  int pad_h = size_[1] - resize_h;
+  if (is_mini_pad_) {
+    pad_h = pad_h % stride_;
+    pad_w = pad_w % stride_;
+  } else if (is_no_pad_) {
+    pad_h = 0;
+    pad_w = 0;
+    resize_h = size_[1];
+    resize_w = size_[0];
+  }
+  if (std::fabs(scale - 1.0f) > 1e-06) {
+    Resize::Run(mat, resize_w, resize_h);
+  }
+  if (pad_h > 0 || pad_w > 0) {
+    float half_h = pad_h * 1.0 / 2;
+    int top = int(round(half_h - 0.1));
+    int bottom = int(round(half_h + 0.1));
+    float half_w = pad_w * 1.0 / 2;
+    int left = int(round(half_w - 0.1));
+    int right = int(round(half_w + 0.1));
+    Pad::Run(mat, top, bottom, left, right, padding_value_);
+  }
+}
+
+bool YOLOv5SegPreprocessor::Preprocess(FDMat* mat, FDTensor* output,
+            std::map<std::string, std::array<float, 2>>* im_info) {
+  // Record the shape of image and the shape of preprocessed image
+  (*im_info)["input_shape"] = {static_cast<float>(mat->Height()),
+                               static_cast<float>(mat->Width())};
+  // yolov5seg's preprocess steps
+  // 1. letterbox
+  // 2. convert_and_permute(swap_rb=true)
+  LetterBox(mat);
+  std::vector<float> alpha = {1.0f / 255.0f, 1.0f / 255.0f, 1.0f / 255.0f};
+  std::vector<float> beta = {0.0f, 0.0f, 0.0f};
+  ConvertAndPermute::Run(mat, alpha, beta, true);
+
+  // Record output shape of preprocessed image
+  (*im_info)["output_shape"] = {static_cast<float>(mat->Height()),
+                                static_cast<float>(mat->Width())};
+
+  mat->ShareWithTensor(output);
+  output->ExpandDim(0);  // reshape to n, h, w, c
+  return true;
+}
+
+bool YOLOv5SegPreprocessor::Run(std::vector<FDMat>* images, std::vector<FDTensor>* outputs,
+                             std::vector<std::map<std::string, std::array<float, 2>>>* ims_info) {
+  if (images->size() == 0) {
+    FDERROR << "The size of input images should be greater than 0." << std::endl;
+    return false;
+  }
+  ims_info->resize(images->size());
+  outputs->resize(1);
+  // Concat all the preprocessed data to a batch tensor
+  std::vector<FDTensor> tensors(images->size()); 
+  for (size_t i = 0; i < images->size(); ++i) {
+    if (!Preprocess(&(*images)[i], &tensors[i], &(*ims_info)[i])) {
+      FDERROR << "Failed to preprocess input image." << std::endl;
+      return false;
+    }
+  }
+
+  if (tensors.size() == 1) {
+    (*outputs)[0] = std::move(tensors[0]);
+  } else {
+    function::Concat(tensors, &((*outputs)[0]), 0);
+  }
+  return true;
+}
+
+}  // namespace detection
+}  // namespace vision
+}  // namespace fastdeploy