Pose Prediction using OpenPose
Action Recognition using pre-trained ST_GCN and MMAction2
Run this code to download the 12 videos for 3 classes of actions: sitting, standing, and sleeping. All videos in the dataset were found on public sources.
The demo code for generating pose predictions using OpenPose can be found in this folder. The notebook installs and uses the OpenPose model trained on BODY_25.
The demo code for generating action predictions for each video can be found in this folder. The notebook installs the MMAction2 toolkit and utilizes their demo_skeleton.py file for generating predictions on each of our twelve videos.