DUBURI AUV Vision Stack

Problem

Autonomous underwater navigation for RoboSub tasks has to run in real time on embedded hardware under poor, shifting underwater visibility — commodity vision pipelines assume clean, well-lit input.

Solution

Built Python/OpenCV pipelines for underwater gate recognition and object detection tuned for embedded constraints, integrated with an Arduino-based control platform for BRAC University's DUBURI AUV team competing in the international RoboSub competition.