JRKagumba / 2D-video-pose-estimation-yolov7

Computer-vision based monocular human pose-estimation and biomechanical gait analysis on elite runners. Using Yolov7 pose-estimation algorithm.

YOLOv7 2D Video Pose Estimation

Overview

This project project is an attempt at a refined version of the NFL Combine - 2D Video Running Analysis. There were issues with defining the specific subject to perform inference on within each frame of video. As well as issues with performing inference with OpenPose. This project aims to resolve both of these issues by using YOLOv7 Pose Estimation algorithm as well as video datasets with only 1 person running in each frame.

In addition to gait analysis, this project aims to combine a previous project of mine Shoe Image Classification with gait analysis. The end goal is still slightly unclear at the time of writing this, but it may be have to do with correlating partucular gait patterns with shoe types and perhaps generating some sort of recommender system based on gait pattern and shoes.

Goals

1. Develop a working model
2. Deploy a web application to allow for live interaction
3. Configure a mobile application to allow for real-time gait event detection

Datasets

The project uses publicly available running footage found from YouTube. At the time of writing this, the datasets consist of 29 treadmill running clips and 50 on ground running clips mainly filmed from a saggital view.

A visual quantitative summary of the "On-Ground" Running Dataset is available here

running_sample.mp4

About YOLOv7 Pose

"You Only Look Once" Pose Estimation

Keypoint Map:

• (5,7,9) ==> Right Arm
• (6,8,10) ==> Left Arm
• (11,13,15) ==> Right Leg
• (12,14,16) ==> Left Leg

Pose estimation implimentation is based on YOLO-Pose.

Dataset preparison

[Keypoints Labels of MS COCO 2017]

Citation

@article{wang2022yolov7,
  title={{YOLOv7}: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors},
  author={Wang, Chien-Yao and Bochkovskiy, Alexey and Liao, Hong-Yuan Mark},
  journal={arXiv preprint arXiv:2207.02696},
  year={2022}
}

Acknowledgements

Expand

• https://github.com/AlexeyAB/darknet
• https://github.com/WongKinYiu/yolor
• https://github.com/WongKinYiu/PyTorch_YOLOv4
• https://github.com/WongKinYiu/ScaledYOLOv4
• https://github.com/Megvii-BaseDetection/YOLOX
• https://github.com/ultralytics/yolov3
• https://github.com/ultralytics/yolov5
• https://github.com/DingXiaoH/RepVGG
• https://github.com/JUGGHM/OREPA_CVPR2022
• https://github.com/TexasInstruments/edgeai-yolov5/tree/yolo-pose