Real-time ISL Recognition on Smart Glasses

Tech stack

• Computer Vision
• Pose Estimation
• Hand Landmarks
• Python
• Real-time Inference

The problem

The deaf and hard-of-hearing community lacks unobtrusive, real-time tools to bridge sign language into everyday interactions. Recognizing Indian Sign Language on a wearable, with low enough latency to feel live, is an open challenge.

The solution

A recognition system built for Ray-Ban Meta smart glasses that combines pose estimation and hand-landmark extraction with a low-latency inference path, so signs are interpreted in real time from the wearer's point of view.

Architecture

Smart-glasses capture → pose estimation + hand-landmark extraction → temporal sign model → low-latency inference. PLACEHOLDER — expand with your model/latency details.

Highlights

• Focused on accessibility for the deaf community
• Combines pose estimation, hand-landmark extraction, and low-latency inference
• Research in progress at IIIT Sri City