APS360 Team 35 - ASL Gesture Recognition

Overview

This project focuses on Sign Language Gesture Recognition using deep learning techniques.

Project Setup

Prerequisites

• Python 3.8+
• Anaconda (for environment management)
• GPU recommended (but not required)

Installation

1. Clone the repository

git clone <repo-url>
cd aps360-team35

2. Set up the environment

Without GPU:

setup.bat
conda activate aps360

With GPU:

gpu_setup.bat  # May require driver updates

3. Run the project

python main.py

Dataset Preparation

Model Storage

• CNN Model: results/cnn/YYYYMMDD_HHMMSS/models/cnn_model.pth
• SVM Model: results/svm/YYYYMMDD_HHMMSS/models/svm_model.pkl
• Scaler:
  • CNN: results/cnn/YYYYMMDD_HHMMSS/models/cnn_scaler.pkl
  • SVM: results/svm/YYYYMMDD_HHMMSS/models/svm_scaler.pkl

Training Artifacts

• CNN Training Plots: results/cnn/YYYYMMDD_HHMMSS/plots/cnn_training_progress.png
• SVM Parameter Search Plots: results/svm/YYYYMMDD_HHMMSS/plots/svm_parameter_search.png
• SVM Learning Curves: results/svm/YYYYMMDD_HHMMSS/plots/svm_learning_curves.png

Metrics and Logs

• CNN Training Results: results/cnn/YYYYMMDD_HHMMSS/logs/cnn_training_results.txt
• SVM Training Results: results/svm/YYYYMMDD_HHMMSS/logs/svm_training_results.txt
• Metrics CSV:
  • CNN: results/cnn/YYYYMMDD_HHMMSS/logs/cnn_metrics.csv
  • SVM: results/svm/YYYYMMDD_HHMMSS/logs/svm_metrics.csv

Summary Files

• CNN Summary: results/cnn/YYYYMMDD_HHMMSS/cnn_summary.json
• SVM Summary: results/svm/YYYYMMDD_HHMMSS/svm_summary.json

Note: YYYYMMDD_HHMMSS represents the timestamp when the model was trained (e.g., 20250306_181739)

Project Structure

sign-language-gesture-recognition/
│
├── data/
│   └── wlasl_data/
│       ├── WLASL_v0.3.json
│       └── video_files/
│
├── src/
│   ├── __init__.py
│   ├── dataset.py          # Contains GestureDataset class and prepare_dataset/download_wlasl_data functions
│   ├── cnn_model.py        # Contains GestureCNN class for CNN implementation
│   ├── svm_model.py        # Contains GestureSVM class for SVM implementation
│   └── train.py            # Contains train_model and train_svm_model functions
│
├── results/
│   ├── cnn/                # Results specific to CNN model (created by main.py)
│   │   └── YYYYMMDD_HHMMSS/  # Timestamp folders for each run
│   │       ├── models/     # Saved model files
│   │       ├── logs/       # Training logs and metrics
│   │       └── plots/      # Visualizations of training progress
│   │
│   ├── svm/                # Results specific to SVM model (created by main.py)
│   │   └── YYYYMMDD_HHMMSS/  # Timestamp folders for each run
│   │       ├── models/     # Saved model files
│   │       ├── logs/       # Training logs and metrics
│   │       └── plots/      # Visualizations and parameter search results
│   │
│   └── unknown/            # For any other model types (created by main.py)
│
├── requirements.txt
├── README.md
└── main.py                 # Main script with save_results function and main() function

---