🤟 Real-Time ASL Detection using Deep Learning 🎥🧠 This project implements a real-time American Sign Language (ASL) alphabet recognition system using a custom-trained deep learning model with OpenCV and TensorFlow/Keras. The model was trained on a dataset of 3000 images per class (A-Z), resized to 200x200 pixels for optimal performance. DataSet: https://www.kaggle.com/datasets
🚀 Features 📷 Live camera detection of ASL hand signs.
🧠 Custom CNN model trained on a high-volume dataset (A1–A3000, B1–B3000, ..., Z1–Z3000).
🗂️ No dependency on text label files – label mapping is handled directly in the code.
⚡ Real-time feedback with frame annotation.
🛠️ Built using TensorFlow, Keras, OpenCV, and NumPy.
🏗️ Dataset Structure css Copy Edit asl_alphabet_train/ ├── A/ │ ├── A1.jpg │ ├── ... │ └── A3000.jpg ├── B/ │ └── ... └── Z/ └── Z3000.jpg 🔧 Tech Stack Python
OpenCV
TensorFlow / Keras
NumPy
🧪 Model Training Model trained on 200x200 color images with 26 output classes (A-Z), using a Convolutional Neural Network with dropout for generalization.
🎯 How to Run Train the model (optional, model already provided).
Run cam.py to activate your webcam and start ASL prediction.
Press Q to quit.
📁 Folder Structure Copy Edit project/ ├── model/ │ └── asl_model.h5 ├── cam.py ├── train_model.py └── ...!(https://github.com/user-attachments/assets/033f99c2-9bcd-48fe-bbbc-140546fa4d3f)
💡 Future Work Add support for dynamic gestures (e.g., "hello", "thank you").
Integrate voice output or subtitles.
Deploy as a web or mobile app.