About the Project

Making Pakistan Sign Language Accessible to Everyone

This system uses computer vision and machine learning to detect and translate Pakistan Sign Language (PSL) gestures in real-time — helping bridge the communication gap between the deaf and hearing communities.

Real-Time Detection

Webcam-based detection runs at up to 30 FPS, recognising hand gestures with sub-second latency using MediaPipe hand landmarks.

MLP Neural Network

A Multi-Layer Perceptron trained on 42 landmark coordinates achieves 98.6% test accuracy across 37 Urdu letters.

Urdu Speech Output

Detected signs are spoken aloud in Urdu using pre-recorded audio, making the system accessible without reading.

Technology Stack
MediaPipe HandsTensorFlow / KerasFastAPINext.js 14OpenCVSQLiteStandardScalerPython
98.6%
Test Accuracy
37
Urdu Letters
42
Landmark Features
16,745
Training Samples