Due to the size limitation on Github, the two demonstrative videos are uploaded to Youtube with the following hyperlinks for reviewing.

Demonstration Video 1. Training and testing the DL model (11 minutes)

Demonstration Video 2. Model deployment for human action recognition (5 minutes)

This respository is a modified version of Microsoft-computervision-recipes library. To install the repository and its dependencies follow these simple steps:

1. (optional) Install Anaconda with Python >= 3.6. Anaconda.

2. Clone the repository (require my admin login)

   git clone https://github.com/TszYiYang/Action_Recognition_R2plus1D-ig65m.git 
   

3. Download the repository as zip file, extract the file to local machine (recommended, no password required)

   Use drop down menu to download the zipped repository
   

4. Install the conda environment, you'll find the environment.yml file in the root directory. To build the conda environment:

   cd Action_Recognition_R2plus1D-ig65m
   conda env create -f environment.yml
   

5. Activate the conda environment and register it with Jupyter:

   conda activate csce636-cv
   python -m ipykernel install --user --name csce636-cv --display-name "Python (csce636-cv)"
   

6. Start the Jupyter notebook server

   jupyter notebook