ASL Alphabet Translator Model is a Deep Learning project designed to recognize and translate American Sign Language (ASL) alphabets from images or real-time video streams. The model utilizes Convolutional Neural Networks (CNN) to classify hand gestures into their corresponding letters (A-Z) along with generic gestures like 'space' or 'del'.
This project aims to bridge the communication gap by providing a digital tool to interpret sign language efficiently using GPU-accelerated computing.
- High Accuracy: Trained on a comprehensive dataset of ASL hand gestures.
- Real-time Prediction: Capable of predicting gestures from a webcam feed.
- GPU Acceleration: Optimized for NVIDIA GPUs using CUDA 11.2 and cuDNN 8.1.
- Visualizations: Includes training metrics (Accuracy & Loss graphs).
The model was trained using the ASL Alphabet Dataset.
- Training Images: 87,000 images (29 classes)
- Image Size: 200x200 pixels
- Classes: A-Z, space, del, nothing
To run this project, please ensure your environment matches the specific versions below to avoid compatibility issues with TensorFlow and GPU drivers:
- Python: 3.10
- TensorFlow: 2.10.0
- CUDA Toolkit: 11.2
- cuDNN: 8.1
Clone the repository
git clone [https://github.com/Ikhsaaan334/ASL-Alphabet-Translator-Model.git](https://github.com/Ikhsaaan334/ASL-Alphabet-Translator-Model.git)
cd ASL-Alphabet-Translator-Model
Create a Virtual Environment (Optional but recommended)
python -m venv venv
# Windows
venv\Scripts\activate
# Mac/Linux
source venv/bin/activate
Training the Model
To train the model from scratch, run the training script
python main.py
Testing / Prediction
python predict.py
Model Performance
Current model metrics:
