VisionAssist

Assistive Object Detection for the Visually Impaired

FPS:0
Inference:0ms
Model:Loading...

Camera not active

Click "Activate Camera" to begin

Controls

50%

About

VisionAssist uses TensorFlow.js and the YOLOv8n model to detect objects in real-time. All processing happens in your browser - no data is sent to any server.

Detects 80 common objects including people, vehicles, animals, and household items with superior accuracy (37% mAP vs COCO-SSD's 22% mAP).

Detection Tips:

  • YOLOv8 excels at detecting small objects (phones, keys, utensils)
  • Lower the confidence threshold to 30-40% for even better small object detection
  • Ensure good lighting and objects are fully visible