/02 · Marine ecosystem / Edge AI

Intelligence, below the surface.

An underwater vision system for detecting mucilage, ghost nets, marine litter, and invasive species — without a cloud connection.

2nd place, TÜBİTAK 2204-A, İzmir Region.

01 / The question

How can underwater monitoring recognize several ecological threats in one system, without relying on the cloud?

02 / My contribution

AI model training and edge-device optimization / quantization.

03 / Technical approach

Edge-AI underwater system detecting mucilage, ghost nets, marine litter, and invasive species in one architecture — no cloud required.
INT8-quantized YOLOv11 pipeline at ~15–30 ms/frame (30–60 FPS) on a Raspberry Pi 5 + Coral TPU, deployable via a single PoE cable.
Role: AI model training and edge-device optimization/quantization.

04 / Outcome & recognition

2nd place, TÜBİTAK 2204-A, İzmir Region.