# Enhanced CenterTrack for Robust Underwater Multi-Fish Tracking

## TL;DR

This paper introduces an improved tracking system for monitoring multiple fish in underwater environments, achieving high accuracy and reliability.

## Contribution

The study proposes a CenterTrack-based framework with three complementary components for robust underwater multi-fish tracking.

## Key findings

- The proposed method achieves an IDF1 score of 82.5% on the MF25 dataset.
- The framework improves trajectory continuity under occlusions and abrupt motion.
- The system is computationally efficient and suitable for real-time tracking in aquaculture.

## Abstract

Tracking multiple fish in underwater environments is essential for studying fish behavior and supporting ecological monitoring. However, poor visibility, complex backgrounds, and frequent occlusions make reliable tracking difficult in real underwater scenes. In this study, we propose an improved method for underwater multi-fish tracking based on CenterTrack. The proposed approach improves tracking stability and accuracy under challenging underwater conditions. Experimental results on underwater fish datasets show that our method performs more reliably than existing approaches. This study provides a practical tool for automated analysis of fish behavior in natural underwater environments.

Accurate monitoring of fish movement is essential for understanding behavioral patterns and group dynamics in aquaculture systems. Underwater scenes—characterized by dense populations, frequent occlusions, non-rigid body motion, and visually similar appearances—present substantial challenges for conventional multi-object tracking methods. We propose an improved CenterTrack-based framework tailored for multi-fish tracking in such environments. The framework integrates three complementary components: a multi-branch feature extractor that enhances discrimination among visually similar individuals, occlusion-aware output heads that estimate visibility states, and a three-stage cascade association module that improves trajectory continuity under abrupt motion and occlusions. To support systematic evaluation, we introduce a self-built dataset named Multi-Fish 25 (MF25), continuous video sequences of 75 individually annotated fish recorded in aquaculture tanks. The experimental results on MF25 show that the proposed method achieves an IDF1 of 82.5%, MOTA of 85.8%, and IDP of 84.7%. Although this study focuses on tracking performance rather than biological analysis, the produced high-quality trajectories form a solid basis for subsequent behavioral studies. The framework’s modular design and computational efficiency make it suitable for practical, online tracking in aquaculture scenarios.