Video Classification System using CNN-LSTM

This project implements a video classification system leveraging a combination of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, focusing on the classification of video data into predefined categories.

Getting Started

Requirements

pip install -r requirements

Configuration

Edit the params.json file to specify the model and data processing parameters. The parameters include paths to data, model configurations, and training settings.

Data Organization

The project expects video data to be organized in a specific structure:

• The data directory should contain subdirectories, each representing a video class.
• Each subdirectory should contain .avi files corresponding to that class.
• The name of each subdirectory is used as the label for the videos inside it.

Example Structure:

Usage

To run the video classification system:

1. Set up the params.json with the desired configuration.
2. Organize your video data as described above.
3. Execute the main script to start the training and classification process.

python main.py

Demo

You can use the demo.py script to see the system in action. Simply modify the labels in the activity_labels with your own data and provide an instance like demo.mpt (which is me doing jumping jacks, and as you can see, the model correctly predicts it).

For more information about the model training process and results, please refer to the report.pdf.