← All work

Case study 03 · 2024 · Underwater robotics

Coral Identification & Tracking for AUVs

Detecting and tracking six coral families in underwater video, so autonomous underwater vehicles can monitor reefs.

Year
2024
Field
Underwater robotics
Stack
YOLOv8, ByteTrack, OpenCV, PyTorch
Result
97.3%mAP@0.5 · 96.6% recall

Overview

A coral identification and tracking pipeline for autonomous underwater vehicles, built with the National Institute of Ocean Technology (NIOT), Chennai.

YOLOv8 was trained on 1,400 underwater frames from the Andaman Islands covering six coral families. Colour correction, white balancing and contrast enhancement make the coral easier to see, and ByteTrack follows each one from frame to frame.

The detector reaches 97.3% mAP at an IoU threshold of 0.5, with 96.6% recall. mAP (mean average precision) scores how accurately a model finds and labels objects. Higher is better.

Published in Current Science, Vol. 128, No. 7 (April 2025): “AI-based channel infused coral identification and tracking algorithm for autonomous underwater vehicles.”