Realtime-Action-Recognition / Realtime-Action-Recognition

A handwash monitoring application using the real-time action recognition algorithm.

This repository has no license, therefore all rights are reserved. You cannot modily or redistribute this code without explicit permission from the copyright holder.

Please send an e-mail to: [e-mail] for any queries.

Real-time Action Recognition System

This is a specific use-case of the Real-time Action Recognition System. In this implementation, we use the new model to demostrate handwash auditing, by using a live feed of a camera setup above a sink where the handwash is being performed.

Instructions to run the code:

1. Install the requirements:

a. In the parent directory, run:

  $ pip install -r requirements.txt

2. Compile and setup PyFlow:

a. Change directory to pyflow

  $ cd pyflow

b. Setup pyflow:

  $ python setup.py build_ext -i

3. Dowload the trained model

a. Download the model from here.

b. Save the model in the parent folder of this repository, with the name: current_final_handwash_model.h5

4. Run the Flask app:

  $ python app.py

5. Open localhost:8001 on your browser when the * Debugger pin is active message appears on the Terminal.

Training and Testing Data:

The training and testing data used was the first split of the UCF-101 Dataset. Due to computation limitations, the first split of the UCF-101 dataset has been preprocessed accoring to our implementation, and stored in the following locations as .pyc files:

The pre-processed training data can be found here.

The pre-processed testing data can be found here.

A Kaggle dataset for this application can be found here.

Points to remember:

1. A webcam must be connected to the computer/laptop.
2. The model must be saved in this repository's parent folder.
3. The computer must have internet access, since JavaScript is dynamically linked.