This repo enables detection of keypoints with the help of Mediapipe API. For conveince, an OpenCV script is implemented to create a dataset quickly. Acquired coordinates are fed into a LSTM network. Base module enables inference on trained actions.
The model is trained for enabling worksite health and safety by detecting actions like waving stop, removing helmet or mask.