03 / 14 — Computer Vision Suite
Sign Language Detection YOLOv8
Real-time sign language gesture detection built on YOLOv8, tuned for edge deployment.
sign-language-detection-yolov8.build
Overview
What it does
Real-time sign language gesture detection built on YOLOv8, tuned for edge deployment.
The Problem
Why it needed building
Real-time sign language recognition needs both high accuracy and low latency — most academic models hit one at the cost of the other, which makes them unusable for live conversation.
The Approach
How it was engineered
Trained and tuned a YOLOv8 detector for hand-gesture classes, then exported through ONNX Runtime to get a model light enough to run on edge GPU hardware without a cloud round-trip.
The Outcome
Where it landed
The final pipeline holds 97.8% mAP@50 at 60 FPS on an edge GPU, with 35ms end-to-end latency from frame capture to label.