anhtu-phan / mmpose

OpenMMLab Pose Estimation Toolbox and Benchmark.

Human Pose Estimation (HPE) in basketball videos

The objective of this project is simulate the human pose in basketball videos which is the movement of players to 2-dimension image.

How to run

Prerequisites

• Linux or macOS
• Python 3.6+
• Pytorch
  • pip install torch==1.7.0+cu110 torchvision==0.8.1+cu110 torchaudio==0.7.0 -f https://download.pytorch.org/whl/torch_stable.html
• MMDetection
  • pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu110/torch1.7.0/index.html
  • git clone https://github.com/open-mmlab/mmdetection.git
  • cd mmdetection
  • pip install -r requirements/build.txt
  • pip install -v -e .
  • cd ..
• MMPose
  • pip install -r requirements.txt
  • python setup.py develop

Run

sh run.sh ${video_path} ${device}

• video_path is path of basket ball video input
• device: cpu for using MMPose with CPU or cuda if using GPU
• the outputs is in new created output folder