FairMOT: On the Fairness of Detection and Re-Identification in Multiple Object Tracking FairMOT is a novel approach for multi-object tracking (MOT) that addresses the fairness issue between object detection and re-identification (re-ID) tasks. The method is based on the anchor-free object detection architecture CenterNet and aims to achieve a balanced optimization of both tasks. Unlike previous methods that treat detection as the primary task and re-ID as secondary, FairMOT treats both tasks equally, which helps reduce bias and improves tracking performance. The approach introduces several key design changes, including the use of object centers for re-ID feature extraction, which avoids the ambiguity caused by anchors and improves feature alignment. The resulting model achieves high accuracy in both detection and tracking, outperforming state-of-the-art methods on multiple public datasets. FairMOT is evaluated on the MOT Challenge benchmark and ranks first on several datasets. It also achieves additional gains when pre-trained using a single image training method. The model runs at 30 FPS on a single RTX 2080 Ti GPU and provides insights into the relationship between detection and re-ID in MOT. The method addresses three main issues: anchor-based design, feature sharing, and feature dimension. By using a homogeneous network structure and extracting features at object centers, FairMOT effectively mitigates these issues and achieves a good trade-off between detection and re-ID. The approach is simple, efficient, and effective, making it a promising solution for MOT.FairMOT: On the Fairness of Detection and Re-Identification in Multiple Object Tracking FairMOT is a novel approach for multi-object tracking (MOT) that addresses the fairness issue between object detection and re-identification (re-ID) tasks. The method is based on the anchor-free object detection architecture CenterNet and aims to achieve a balanced optimization of both tasks. Unlike previous methods that treat detection as the primary task and re-ID as secondary, FairMOT treats both tasks equally, which helps reduce bias and improves tracking performance. The approach introduces several key design changes, including the use of object centers for re-ID feature extraction, which avoids the ambiguity caused by anchors and improves feature alignment. The resulting model achieves high accuracy in both detection and tracking, outperforming state-of-the-art methods on multiple public datasets. FairMOT is evaluated on the MOT Challenge benchmark and ranks first on several datasets. It also achieves additional gains when pre-trained using a single image training method. The model runs at 30 FPS on a single RTX 2080 Ti GPU and provides insights into the relationship between detection and re-ID in MOT. The method addresses three main issues: anchor-based design, feature sharing, and feature dimension. By using a homogeneous network structure and extracting features at object centers, FairMOT effectively mitigates these issues and achieves a good trade-off between detection and re-ID. The approach is simple, efficient, and effective, making it a promising solution for MOT.