zishuo/ascend_community_projects
1
0
Fork 0
mirror of https://github.com/Ascend/ascend_community_projects.git synced 2025-09-26 20:01:17 +08:00
Code Issues Packages Projects Releases Wiki Activity

Modified some normative problems and visualization part of the code

Browse Source
This commit is contained in:
warmy1
2022-06-14 08:49:27 +00:00
parent 32bb962553
commit 491c7b8a05
10 changed files with 239 additions and 188 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
AlphaPose/README.md
View File

@@ -136,10 +136,10 @@ export PYTHONPATH=${MX_SDK_HOME}/python:$PYTHONPATH
推理中涉及到第三方软件依赖如下表所示。
| 依赖软件 | 版本 | 说明 | 使用教程 |
| -------- | ---------- | ------------------------------ | ------------------------------------------------------------ |
| live555 | 1.09 | 实现视频转rstp进行推流 | [链接](https://gitee.com/ascend/mindxsdk-referenceapps/blob/master/docs/%E5%8F%82%E8%80%83%E8%B5%84%E6%96%99/Live555%E7%A6%BB%E7%BA%BF%E8%A7%86%E9%A2%91%E8%BD%ACRTSP%E8%AF%B4%E6%98%8E%E6%96%87%E6%A1%A3.md) |
| ffmpeg | 2021-07-21 | 实现mp4格式视频转为264格式视频 | [链接](https://gitee.com/ascend/mindxsdk-referenceapps/blob/master/docs/%E5%8F%82%E8%80%83%E8%B5%84%E6%96%99/pc%E7%AB%AFffmpeg%E5%AE%89%E8%A3%85%E6%95%99%E7%A8%8B.md#https://ffmpeg.org/download.html) |
| 依赖软件 | 版本 | 说明 | 使用教程 |
| -------- | ---------- | ------------------------------ | ------------------------------------------------------- |
| live555 | 1.09 | 实现视频转rstp进行推流 | [链接](https://gitee.com/ascend/mindxsdk-referenceapps) |
| ffmpeg | 2021-07-21 | 实现mp4格式视频转为264格式视频 | [链接](https://gitee.com/ascend/mindxsdk-referenceapps) |
@@ -191,7 +191,7 @@ ATC run success, welcome to the next use.
## 5 准备
按照第 3 小结**软件依赖**安装 live555 和 ffmpeg，按照 [Live555离线视频转RTSP说明文档 ](https://gitee.com/ascend/mindxsdk-referenceapps/blob/master/docs/%E5%8F%82%E8%80%83%E8%B5%84%E6%96%99/Live555%E7%A6%BB%E7%BA%BF%E8%A7%86%E9%A2%91%E8%BD%ACRTSP%E8%AF%B4%E6%98%8E%E6%96%87%E6%A1%A3.md)将 mp4 视频转换为 h264 格式。并将生成的 264 格式的视频上传到 `live/mediaServer` 目录下，然后修改 `AlphaPose/pipeline` 目录下的 `video.pipeline` 文件中 mxpi_rtspsrc0 的内容。
按照第 3 小结**软件依赖**安装 live555 和 ffmpeg，按照 [Live555离线视频转RTSP说明文档 ](https://gitee.com/ascend/mindxsdk-referenceapps)将 mp4 视频转换为 h264 格式。并将生成的 264 格式的视频上传到 `live/mediaServer` 目录下，然后修改 `AlphaPose/pipeline` 目录下的 `video.pipeline` 文件中 mxpi_rtspsrc0 的内容。
```
"mxpi_rtspsrc0": {
@@ -248,9 +248,7 @@ bash run.sh image
bash run.sh video --speedtest
```
结果将如下图所示，终端会每 10 帧打印一次当前帧数，前 10 帧的运行时间以及前10帧的平均帧率。
![性能测试结果](./image/speedtest.png)
命令运行后终端会每 10 帧打印一次当前帧数，前 10 帧的运行时间以及前10帧的平均帧率。经测试，性能可以达到 19 fps 左右，满足性能要求。
[^注]: 输入视频帧率应高于25，否则无法发挥全部性能。且由于 Alphapose 人体关键点估计是一种自上而下的方式，所以实际推理速度与视频中的人数存在负相关关系，即人数越多，推理用时越多，速度越慢。上述展示的推理速度是在视频帧大小为 720*1280，且视频中只有一个人的条件下所得到的性能。
@@ -278,7 +276,7 @@ bash run.sh video --speedtest
└── video.py
```
3. 执行第 6 小节 **编译与运行** 中的步骤 1 至步骤 3 完成准备工作，进入 `AlphaPose` 目录，在 `AlphaPose` 目录下执行命令：
@@ -286,9 +284,7 @@ bash run.sh video --speedtest
bash run.sh evaluate
```
命令执行结束后输出 COCO 格式的评测结果，并生成 val2017_keypoint_detect_result.json 检测结果文件。输出结果如下图所示：
![精度测试结果](./image/acctest.png)
命令执行结束后输出 COCO 格式的评测结果，并生成 val2017_keypoint_detect_result.json 检测结果文件。经测试，在coco验证集上的 mAP 为 0.732，满足精度要求。
## 8 常见问题

2
AlphaPose/build.sh
View File

@@ -1,5 +1,5 @@
#!/bin/bash
# Copyright 2021 Huawei Technologies Co., Ltd
# Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.

104
AlphaPose/plugin/postprocess/MxpiAlphaposePostProcess.cpp
View File

@@ -398,6 +398,50 @@ static void PoseMergeFast(const cv::Mat &predsPick, const std::vector<cv::Mat> &
}
}
/**
* @brief Delete the selected person
* @param tmpSize - the number of person now
* @param pickId - The person id picked
* @param keypointPreds - Source data containing the information of keypoints position
* @param keypointScores - Source data containing the information of keypoins score
* @param humanIds - A collection of person ids
* @param humanScores - A collection of the information of person's score
* @param confidence - A collection of the confidece of person
* @param mergeIds - A collection of the pose to merge
*/
static void DeletePerson(int tmpSize, int pickId,
std::vector<float> &finalDists, std::vector<int> &numMatchKeypoints,
std::vector<cv::Mat> &keypointPreds, std::vector<cv::Mat> &keypointScores,
std::vector<int> &humanIds, std::vector<float> &humanScores,
std::vector<float> &confidence, std::vector<int> &mergeIds)
{
int count = 0;
float gma = 22.48;
for (int i = 0; i < tmpSize; i++) {
if ((finalDists[i] > gma)||(numMatchKeypoints[i] > MATCH_THREAD)) {
int deleteId = i - count;
int mergeId = humanIds[deleteId];
mergeIds.push_back(mergeId);
keypointPreds.erase(keypointPreds.begin() + deleteId);
keypointScores.erase(keypointScores.begin() + deleteId);
humanIds.erase(humanIds.begin() + deleteId);
humanScores.erase(humanScores.begin() + deleteId);
confidence.erase(confidence.begin() + deleteId);
count ++;
}
}
if (mergeIds.size() == 0) {
int deleteId = pickId;
int mergeId = humanIds[deleteId];
mergeIds.push_back(mergeId);
keypointPreds.erase(keypointPreds.begin() + deleteId);
keypointScores.erase(keypointScores.begin() + deleteId);
humanIds.erase(humanIds.begin() + deleteId);
humanScores.erase(humanScores.begin() + deleteId);
confidence.erase(confidence.begin() + deleteId);
}
}
/**
* @brief Get final pose
* @param maxValue - The max value of picked object's keypoints scores
@@ -499,24 +543,15 @@ APP_ERROR MxpiAlphaposePostProcess::ExtractKeypointsInfo(const std::vector<std::
* @param personScores - Target data containing the information of person's score
* @return APP_ERROR
*/
APP_ERROR MxpiAlphaposePostProcess::PoseNms(std::vector<cv::Mat> &keypointPreds,
std::vector<cv::Mat> &keypointScores,
std::vector<std::vector<float> > &objectBoxes,
std::vector<cv::Mat> &finalPoses,
std::vector<cv::Mat> &finalScores,
std::vector<float> &personScores)
APP_ERROR MxpiAlphaposePostProcess::PoseNms(std::vector<cv::Mat> &keypointPreds, std::vector<cv::Mat> &keypointScores,
std::vector<std::vector<float> > &objectBoxes, std::vector<cv::Mat> &finalPoses,
std::vector<cv::Mat> &finalScores, std::vector<float> &personScores)
{
float gma = 22.48;
float alpha = 0.1;
int batchNum = keypointScores.size();
std::vector<float> confidence(batchNum);
std::vector<float> boxWidth(batchNum);
std::vector<float> boxHeight(batchNum);
std::vector<float> refDists(batchNum);
std::vector<float> humanScores(batchNum);
std::vector<int> humanIds(batchNum);
std::vector<float> finalDists(batchNum);
std::vector<int> numMatchKeypoints(batchNum);
std::vector<float> confidence(batchNum), boxWidth(batchNum), boxHeight(batchNum);
std::vector<float> refDists(batchNum), humanScores(batchNum), finalDists(batchNum);
std::vector<int> humanIds(batchNum), numMatchKeypoints(batchNum);
for (int i = 0; i < batchNum; i++) {
for (int j = 0; j < keypointScores[i].rows; j++) {
float *ptr = keypointScores[i].ptr<float>(j);
@@ -554,47 +589,16 @@ APP_ERROR MxpiAlphaposePostProcess::PoseNms(std::vector<cv::Mat> &keypointPreds,
PCKMatch(pickId, refDist, keypointPreds, numMatchKeypoints);
// Delete humans who have more than MATCH_THREAD keypoints overlap and high similarity
std::vector<int> mergeIds = {};
std::vector<int> deleteIds = {};
int count = 0;
for (int i = 0; i < tmpSize; i++) {
if ((finalDists[i] > gma)||(numMatchKeypoints[i] > MATCH_THREAD)) {
int deleteId = i - count;
int mergeId = humanIds[deleteId];
mergeIds.push_back(mergeId);
keypointPreds.erase(keypointPreds.begin() + deleteId);
keypointScores.erase(keypointScores.begin() + deleteId);
humanIds.erase(humanIds.begin() + deleteId);
humanScores.erase(humanScores.begin() + deleteId);
confidence.erase(confidence.begin() + deleteId);
count ++;
}
}
if (mergeIds.size() == 0) {
int deleteId = pickId;
int mergeId = humanIds[deleteId];
mergeIds.push_back(mergeId);
keypointPreds.erase(keypointPreds.begin() + deleteId);
keypointScores.erase(keypointScores.begin() + deleteId);
humanIds.erase(humanIds.begin() + deleteId);
humanScores.erase(humanScores.begin() + deleteId);
confidence.erase(confidence.begin() + deleteId);
}
DeletePerson(tmpSize, pickId, finalDists, numMatchKeypoints, keypointPreds, keypointScores,
humanIds, humanScores, confidence, mergeIds);
double maxValue;
cv::Point maxIdx;
cv::minMaxLoc(scoresPick, NULL, &maxValue, NULL, &maxIdx);
if (maxValue >= SCORE_THREAD) {
cv::Mat mergePose(KEY_POINTS_NUM, POSE_COORD_NUM, CV_32FC1, Scalar(0));
cv::Mat mergeScore(KEY_POINTS_NUM, SCORE_COORD_NUM, CV_32FC1, Scalar(0));
if (mergeIds.size() == 1) {
mergePose = originKeypointPreds[mergeIds[0]].clone();
mergeScore = originKeypointScores[mergeIds[0]].clone();
} else {
PoseMergeFast(predsPick, originKeypointPreds, originKeypointScores,
refDist, mergeIds, mergePose, mergeScore);
}
GetFinalPose(maxValue, confidencePick, mergePose,
mergeScore, finalPoses, finalScores, personScores);
PoseMergeFast(predsPick, originKeypointPreds, originKeypointScores, refDist, mergeIds, mergePose, mergeScore);
GetFinalPose(maxValue, confidencePick, mergePose, mergeScore, finalPoses, finalScores, personScores);
}
}
return APP_ERR_OK;

10
AlphaPose/plugin/postprocess/build.sh
View File

@@ -1,5 +1,5 @@
#!/bin/bash
# Copyright 2021 Huawei Technologies Co., Ltd
# Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.
@@ -11,9 +11,9 @@
# 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.mitations under the License.
# limitations under the License.
set -e
set -e
current_folder="$( cd "$(dirname "$0")" ;pwd -P )"
@@ -21,7 +21,7 @@ function build_plugin() {
build_path=${current_folder}/build
plugin_path=${MX_SDK_HOME}/lib/plugins
plugin_name="libmxpi_alphaposepostprocess.so"
if [ -d "$build_path" ]; then
rm -rf "$build_path"
else
@@ -31,7 +31,7 @@ function build_plugin() {
cd "$build_path"
cmake ..
make -j4
chmod 440 "$plugin_name"
if [ -e "${plugin_path}/${plugin_name}" ];
then rm -rf "${plugin_path}/${plugin_name}"

4
AlphaPose/plugin/preprocess/build.sh
View File

@@ -1,5 +1,5 @@
#!/bin/bash
# Copyright 2021 Huawei Technologies Co., Ltd
# Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.
@@ -11,7 +11,7 @@
# 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.mitations under the License.
# limitations under the License.
set -e

4
AlphaPose/proto/build.sh
View File

@@ -1,5 +1,5 @@
#!/bin/bash
# Copyright 2021 Huawei Technologies Co., Ltd
# Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.
@@ -11,7 +11,7 @@
# 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.mitations under the License.
# limitations under the License.
set -e

14
AlphaPose/run.sh
View File

@@ -1,3 +1,17 @@
#!/bin/bash
# Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.
set -e
MODE=$1

39
AlphaPose/src/image.py
View File

@@ -1,23 +1,19 @@
#!/usr/bin/env python
# coding=utf-8
"""
Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.
"""
# Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.
import json
import sys
import stat
@@ -84,8 +80,9 @@ def main():
personlist = []
for i, _ in enumerate(range(person_num)):
person = pose_out_list.personInfoVec[i]
keypoints_score = np.zeros((17, 1), dtype = np.float32)
keypoints_pre = np.zeros((17, 2), dtype = np.float32)
keypoints_num = 17
keypoints_score = np.zeros((keypoints_num, 1), dtype = np.float32)
keypoints_pre = np.zeros((keypoints_num, 2), dtype = np.float32)
for j in range(len(person.keyPoints)):
keypoints_score[j][0] = person.keyPoints[j].score
keypoints_pre[j][0] = person.keyPoints[j].x
@@ -93,7 +90,7 @@ def main():
score = np.array(person.confidence)
personlist.append({
'keypoints': keypoints_pre,
'kp_score': keypoints_score,
'keypoints_score': keypoints_score,
'proposal_score': score
})
# Read the original image
@@ -105,7 +102,7 @@ def main():
# Save key point information to JSON file
for i, _ in enumerate(range(person_num)):
personlist[i]['keypoints'] = personlist[i]['keypoints'].tolist()
personlist[i]['kp_score'] = personlist[i]['kp_score'].tolist()
personlist[i]['keypoints_score'] = personlist[i]['keypoints_score'].tolist()
personlist[i]['proposal_score'] = personlist[i]['proposal_score'].tolist()
flags = os.O_WRONLY | os.O_CREAT # Sets how files are read and written
modes = stat.S_IWUSR | stat.S_IRUSR # Set file permissions

189
AlphaPose/src/utils/visualization.py
View File

@@ -1,5 +1,7 @@
# ------------------------------------------------------------------------------
# Copyright 2021 Huawei Technologies Co., Ltd
#!/usr/bin/env python
# coding=utf-8
# Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.
@@ -12,97 +14,138 @@
# 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.
# ============================================================================
import cv2
import math
import numpy as np
import cv2
YELLOW = (0, 255, 255)
ORANGE = (0, 165, 255)
CYAN = (255, 255, 0)
PURPLE = (255, 0, 255)
RED = (0, 0, 255)
GREEN = (0, 255, 0)
BLUE = (255, 0, 0)
CYAN = (255, 255, 0)
YELLOW = (0, 255, 255)
ORANGE = (0, 165, 255)
PURPLE = (255, 0, 255)
def visualize(frame, result, dataset='coco'):
'''
frame: frame image
result: result of predictions
dataset: coco or mpii
return rendered image
'''
if dataset == 'coco':
l_pair = [
(0, 1), (0, 2), (1, 3), (2, 4), # Head
(5, 6), (5, 7), (7, 9), (6, 8), (8, 10),
(17, 11), (17, 12), # Body
(11, 13), (12, 14), (13, 15), (14, 16)
]
p_color = [(0, 255, 255), (0, 191, 255), (0, 255, 102), (0, 77, 255), (0, 255, 0),
(77, 255, 255), (77, 255, 204), (77, 204, 255), (191, 255, 77), (77, 191, 255), (191, 255, 77),
(204, 77, 255), (77, 255, 204), (191, 77, 255), (77, 255, 191),
(127, 77, 255), (77, 255, 127), (0, 255, 255)]
line_color = [(0, 215, 255), (0, 255, 204), (0, 134, 255), (0, 255, 50),
(77, 255, 222), (77, 196, 255), (77, 135, 255), (191, 255, 77), (77, 255, 77),
(77, 222, 255), (255, 156, 127),
(0, 127, 255), (255, 127, 77), (0, 77, 255), (255, 77, 36)]
elif dataset == 'mpii':
l_pair = [
(8, 9), (11, 12), (11, 10), (2, 1), (1, 0),
(13, 14), (14, 15), (3, 4), (4, 5),
(8, 7), (7, 6), (6, 2), (6, 3), (8, 12), (8, 13)
]
p_color = [PURPLE, BLUE, BLUE, RED, RED, BLUE, BLUE, RED, RED, PURPLE, PURPLE, PURPLE, RED, RED, BLUE, BLUE]
line_color = [PURPLE, BLUE, BLUE, RED, RED, BLUE, BLUE, RED, RED, PURPLE, PURPLE, RED, RED, BLUE, BLUE]
else:
raise NotImplementedError
def visualize_fast(frame, result):
line_pair = [
(0, 1), (0, 2), (1, 3), (2, 4),
(5, 6), (5, 7), (7, 9), (6, 8), (8, 10),
(17, 11), (17, 12),
(11, 13), (12, 14), (13, 15), (14, 16)
]
# Nose, LEye, REye, LEar, REar
# LShoulder, RShoulder, LElbow, RElbow, LWrist, RWrist
# LHip, RHip, LKnee, Rknee, LAnkle, RAnkle, Neck
point_color = [(0, 255, 255), (0, 191, 255), (0, 255, 102), (0, 77, 255), (0, 255, 0),
(77, 255, 255), (77, 255, 204), (77, 204, 255), (191, 255, 77), (77, 191, 255), (191, 255, 77),
(204, 77, 255), (77, 255, 204), (191, 77, 255), (77, 255, 191), (127, 77, 255), (77, 255, 127), (0, 255, 255)]
line_color = [(0, 215, 255), (0, 255, 204), (0, 134, 255), (0, 255, 50),
(77, 255, 222), (77, 196, 255), (77, 135, 255), (191, 255, 77), (77, 255, 77),
(77, 222, 255), (255, 156, 127),
(0, 127, 255), (255, 127, 77), (0, 77, 255), (255, 77, 36)]
img = frame
for human in result:
part_line = {}
kp_preds = human['keypoints']
kp_scores = human['kp_score']
kp_preds = np.concatenate((kp_preds, np.expand_dims((kp_preds[5, :]+kp_preds[6, :]) / 2.0, 0)))
kp_scores = np.concatenate((kp_scores, np.expand_dims((kp_scores[5, :]+kp_scores[6, :]) / 2.0, 0)))
preds = human['keypoints']
scores = human['keypoints_score']
# Get the keypoint between the left shoulder and the right shoulder
lshoulder_index = 5
rshoulder_index = 6
middle_preds = np.expand_dims((preds[lshoulder_index, :]+preds[rshoulder_index, :]) / 2.0, 0)
middle_scores = np.expand_dims((scores[lshoulder_index, :]+scores[rshoulder_index, :]) / 2.0, 0)
preds = np.concatenate((preds, middle_preds))
scores = np.concatenate((scores, middle_scores))
# Draw keypoints
for n in range(kp_scores.shape[0]):
if kp_scores[n] <= 0.05:
for n in range(scores.shape[0]):
if scores[n] <= 0.05:
continue
cor_x, cor_y = int(kp_preds[n, 0]), int(kp_preds[n, 1])
part_line[n] = (int(cor_x), int(cor_y))
bg = img.copy()
cv2.circle(bg, (int(cor_x), int(cor_y)), 2, p_color[n], -1)
# Now create a mask of logo and create its inverse mask also
transparency = float(max(0, min(1, kp_scores[n])))
img = cv2.addWeighted(bg, transparency, img, 1-transparency, 0)
for i, (start_pair, end_pair) in enumerate(l_pair):
point_x, point_y = int(preds[n, 0]), int(preds[n, 1])
part_line[n] = (int(point_x), int(point_y))
if n < len(point_color):
cv2.circle(img, (point_x, point_y), 3, point_color[n], -1)
else:
cv2.circle(img, (point_x, point_y), 1, (255, 255, 255), 2)
# Draw limbs
for i, (start_pair, end_pair) in enumerate(line_pair):
if start_pair in part_line and end_pair in part_line:
start_xy = part_line.get(start_pair)
end_xy = part_line.get(end_pair)
bg = img.copy()
start_point = part_line.get(start_pair)
end_point = part_line.get(end_pair)
if i < len(line_color):
cv2.line(img, start_point, end_point, line_color[i],
2 * int(scores[start_pair] + scores[start_pair]) + 1)
else:
cv2.line(img, start_point, end_point, (255, 255, 255), 1)
return img
coord_x = (start_xy[0], end_xy[0])
coord_y = (start_xy[1], end_xy[1])
def visualize(frame, result):
line_pair = [
(0, 1), (0, 2), (1, 3), (2, 4),
(5, 6), (5, 7), (7, 9), (6, 8), (8, 10),
(17, 11), (17, 12),
(11, 13), (12, 14), (13, 15), (14, 16)
]
# Nose, LEye, REye, LEar, REar
# LShoulder, RShoulder, LElbow, RElbow, LWrist, RWrist
# LHip, RHip, LKnee, Rknee, LAnkle, RAnkle, Neck
point_color = [(0, 255, 255), (0, 191, 255), (0, 255, 102), (0, 77, 255), (0, 255, 0),
(77, 255, 255), (77, 255, 204), (77, 204, 255), (191, 255, 77), (77, 191, 255), (191, 255, 77),
(204, 77, 255), (77, 255, 204), (191, 77, 255), (77, 255, 191), (127, 77, 255), (77, 255, 127), (0, 255, 255)]
line_color = [(0, 215, 255), (0, 255, 204), (0, 134, 255), (0, 255, 50),
(77, 255, 222), (77, 196, 255), (77, 135, 255), (191, 255, 77), (77, 255, 77),
(77, 222, 255), (255, 156, 127),
(0, 127, 255), (255, 127, 77), (0, 77, 255), (255, 77, 36)]
img = frame
for human in result:
part_line = {}
preds = human['keypoints']
scores = human['keypoints_score']
# Get the keypoint between the left shoulder and the right shoulder
lshoulder_index = 5
rshoulder_index = 6
middle_preds = np.expand_dims((preds[lshoulder_index, :]+preds[rshoulder_index, :]) / 2.0, 0)
middle_scores = np.expand_dims((scores[lshoulder_index, :]+scores[rshoulder_index, :]) / 2.0, 0)
preds = np.concatenate((preds, middle_preds))
scores = np.concatenate((scores, middle_scores))
# Draw keypoints
for n in range(scores.shape[0]):
if scores[n] <= 0.05:
continue
point_x, point_y = int(preds[n, 0]), int(preds[n, 1])
part_line[n] = (int(point_x), int(point_y))
point_img = img.copy()
cv2.circle(point_img, (int(point_x), int(point_y)), 2, point_color[n], -1)
# Now create a mask of logo and create its inverse mask also
transparency = float(max(0, min(1, scores[n])))
img = cv2.addWeighted(point_img, transparency, img, 1-transparency, 0)
# Draw limbs
for i, (start_pair, end_pair) in enumerate(line_pair):
if start_pair in part_line and end_pair in part_line:
start_point = part_line.get(start_pair)
end_point = part_line.get(end_pair)
line_img = img.copy()
coord_x = (start_point[0], end_point[0])
coord_y = (start_point[1], end_point[1])
coord_mx = np.mean(coord_x)
coord_my = np.mean(coord_y)
length = ((coord_y[0] - coord_y[1]) ** 2 + (coord_x[0] - coord_x[1]) ** 2) ** 0.5
angle = math.degrees(math.atan2(coord_y[0] - coord_y[1], coord_x[0] - coord_x[1]))
stickwidth = (kp_scores[start_pair] + kp_scores[end_pair]) + 1
length = ((start_point[1] - end_point[1]) ** 2 +
(start_point[0] - end_point[0]) ** 2) ** 0.5
angle_degrees = math.atan2(start_point[1] - end_point[1], start_point[0] - end_point[0])
angle = math.degrees(angle_degrees)
stick_width = (scores[start_pair] + scores[end_pair]) + 1
polygon = cv2.ellipse2Poly((int(coord_mx), int(coord_my)), (int(length/2),
int(stickwidth)), int(angle), 0, 360, 1)
int(stick_width)), int(angle), 0, 360, 1)
if i < len(line_color):
cv2.fillConvexPoly(bg, polygon, line_color[i])
cv2.fillConvexPoly(line_img, polygon, line_color[i])
else:
cv2.line(bg, start_xy, end_xy, (255, 255, 255), 1)
if n < len(p_color):
transparency = float(max(0, min(1, 0.5 * (kp_scores[start_pair] + kp_scores[end_pair])-0.1)))
cv2.line(line_img, start_point, end_point, (255, 255, 255), 1)
if n < len(point_color):
transparency = float(max(0, min(1, 0.5 * (scores[start_pair] + scores[end_pair])-0.1)))
else:
transparency = float(max(0, min(1, (kp_scores[start_pair] + kp_scores[end_pair]))))
transparency = float(max(0, min(1, (scores[start_pair] + scores[end_pair]))))
img = cv2.addWeighted(bg, transparency, img, 1 - transparency, 0)
img = cv2.addWeighted(line_img, transparency, img, 1 - transparency, 0)
return img

41
AlphaPose/src/video.py
View File

@@ -1,23 +1,19 @@
#!/usr/bin/env python
# coding=utf-8
"""
Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.
"""
# Copyright(C) 2021. Huawei Technologies Co.,Ltd. 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.
import argparse
import json
import os
@@ -30,7 +26,7 @@ import numpy as np
import MxpiDataType_pb2 as MxpiDataType
from StreamManagerApi import StreamManagerApi, StringVector
from utils.visualization import visualize
from utils.visualization import visualize, visualize_fast
sys.path.append("../proto/")
import mxpiAlphaposeProto_pb2 as mxpialphaposeproto
@@ -102,8 +98,9 @@ def main():
personlist = []
for i, _ in enumerate(range(person_num)):
person = pose_out_list.personInfoVec[i]
keypoints_score = np.zeros((17, 1), dtype = np.float32)
keypoints_pre = np.zeros((17, 2), dtype = np.float32)
keypoints_num = 17
keypoints_score = np.zeros((keypoints_num, 1), dtype = np.float32)
keypoints_pre = np.zeros((keypoints_num, 2), dtype = np.float32)
for j in range(len(person.keyPoints)):
keypoints_score[j][0] = person.keyPoints[j].score
keypoints_pre[j][0] = person.keyPoints[j].x
@@ -112,7 +109,7 @@ def main():
personlist.append({
'frame': frame_count,
'keypoints': keypoints_pre,
'kp_score': keypoints_score,
'keypoints_score': keypoints_score,
'proposal_score': score,
})
# Whether to conduct speed tests
@@ -124,7 +121,7 @@ def main():
# Save key point information to JSON file
for i, _ in enumerate(range(person_num)):
personlist[i]['keypoints'] = personlist[i]['keypoints'].tolist()
personlist[i]['kp_score'] = personlist[i]['kp_score'].tolist()
personlist[i]['keypoints_score'] = personlist[i]['keypoints_score'].tolist()
personlist[i]['proposal_score'] = personlist[i]['proposal_score'].tolist()
flags = os.O_WRONLY | os.O_APPEND | os.O_CREAT # Sets how files are read and written
modes = stat.S_IWUSR | stat.S_IRUSR # Set file permissions