Wearable Conversational Assistance for Blind and Visually Impaired People
Wearable Conversational Assistance for Blind and Visually Impaired People
An ongoing research-oriented project exploring how computer vision can assist blind and visually impaired individuals during social interactions. The objective is to provide environmental awareness and contextual information that can help users better understand their surroundings during conversations.
Technical aspects explored:
- Python-based computer vision pipeline
- OpenCV
- YOLOv8-based face detection
- BoT-SORT tracking
- MediaPipe facial landmarks
- SmolVLM-based scene description
The system explores face tracking and scene understanding to provide contextual visual assistance. Currently a laptop-based prototype with potential Raspberry Pi and camera deployment. This is an ongoing research project, not a finished commercial product.
Note: This project does not currently implement audio-based speaker identification.