DeepBroilerTrack: End-to-end automatic tracking of multiple broilers using multiple cameras

Reliable tracking of individual broilers in commercial poultry facilities is essential for large-scale welfare monitoring, behavior analysis, and precision breeding. Multi-camera systems have emerged to address challenges posed by occlusions, dense flock layouts, and strong appearance similarity. However, existing approaches often rely on hand-crafted association heuristics and cascade-style two-stage designs, which struggle to generalize in crowded and occlusion-heavy environments. To overcome these shortcomings, we propose DeepBroilerTrack, an end-to-end multi-camera tracking framework unifying early-stage feature fusion within a Bird’s-Eye View (BEV) representation. The framework consists of eight key modules: Image Augmentation, Image Encoder, Image Feature Enhancement, Cross-View Aggregation, BEV Detector, BEV Tracker, Image Object Detector, and ID Assignment. The Image Augmentation strategy, combining view geometric transformations and camera dropout, enhances generalization. The Image Encoder extracts initial feature representations, which are refined by the Image Feature Enhancement module. Image features from camera views are fused by the Cross-View Aggregation module to construct a BEV feature, which is then used for the BEV Decoder to extract BEV objects. The detected BEV objects are temporally linked into tracklets via the BEV Tracker module. These BEV tracklets are subsequently projected back into each camera view through the ID Assignment process and aligned with per-view detections from the Image Object Detector module, yielding ID-consistent trajectories across all cameras. Experiments on a real-world broiler dataset demonstrate that our method significantly improves tracking stability and identity (ID) consistency over existing multi-camera baselines, offering a practical and scalable solution for intelligent poultry management in challenging production settings.​