This is the authors' abstract. We don't add key points for this paper.
Multi-object tracking (MOT) in complex environments remains a significant challenge due to occlusions, visual similarities, and the need for long-term identity preservation, particularly in animal behavior studies. This paper presents a unified tracking pipeline for cross-species monitoring of zebrafish and chickens, addressing these issues in aquatic and terrestrial settings. Our approach integrates a customized DeepSORT tracker with YOLOv8 for real-time object detection, which is further improved by Kalman filtering and a ResNet-50-based Re-Identification (ReID) model. With a mean Average Precision (mAP@0.5) of 0.92 for zebrafish detection and 0.563 for chickens, the system demonstrated strong tracking accuracy even in the face of occlusions and overlapping motions when tested on the 3D-ZeF20 dataset for zebrafish and the AnimalTrack dataset for chickens. While motion-based tracking was adequate for hens in controlled settings, the ReID module dramatically decreased identification shifts in visually complex zebrafish environments. This adaptable system provides a scalable solution for realtime monitoring and welfare evaluation across a variety of species, improving automated behavioral analysis in ethology and precision livestock farming.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Aboud et al. (2025) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: