Subfolders named "Dataset_X" contain code to generate training/validation/test csv files containing MoveNet coordinates for the respective datasets.
The "Deep Learning" subfolder contains Jupyter Notebooks to train the various models and some Fisher Ratio utilities
The "InsightFace_Testing" subfolder contains (i) a Streamlit app to test InsightFace face detection with FPS, and (ii) code to test InsightFace with blurring on a given video (it will save the processed video)
The "insightface_vs_proposed" subfolder contains code for (i) OpenCV FaceCascade, (ii) InsightFace face detection, (iii) Proposed face detection method in order to compare them through a proposed overlap metric (overlap_metrics.py)
The "MoveNet_Testing" subfolder contains code to test the classifier system on images (movenet_testing_image.py) and also on a video (movenet_testing_video.py)
The "WebApp" subfolder contains the necessary code to run and deploy a Streamlit web application for the human activity recognition system
For any code using a Streamlit web application, use the app_envm.yml file for the conda environment. For the notebooks in "Deep Learning" subfolder, run these in Google Colab. For other non-app code, use the non_app_envm.yml file for the environment.
@inproceedings{singh2023,
author = {Singh, Ishneet Sukhvinder and Kaza, Pradyoth and Hosler Iv, Peter Gregory and Chin, Zheng Yang and Ang, Kai Keng},
title = {Real-Time Privacy Preserving Human Activity Recognition on Mobile using 1DCNN-BiLSTM Deep Learning},
year = {2023},
booktitle = {Proceedings of the 2023 5th International Conference on Image, Video and Signal Processing},
}