ByteTrack is a novel multi-object tracking (MOT) method that addresses the issue of missing and fragmented trajectories by associating almost every detection box, regardless of its score. Traditional methods often discard low-confidence detection boxes, leading to significant errors in MOT performance. ByteTrack uses a simple and effective association method, BYTE, which associates high-confidence detection boxes with tracklets and then matches low-confidence detection boxes to these tracklets. This approach helps recover occluded objects and filters out background detections. The method is evaluated on nine state-of-the-art trackers, showing consistent improvements in metrics such as MOTA, IDF1, and HOTA. ByteTrack is also designed as a strong tracker, achieving 80.3 MOTA, 77.3 IDF1, and 63.1 HOTA on the MOT17 dataset with 30 FPS on a single V100 GPU. It outperforms other trackers on various benchmarks, including MOT20, HiEve, and BDD100K. The source code and pre-trained models are available at <https://github.com/ifzhang/ByteTrack>.ByteTrack is a novel multi-object tracking (MOT) method that addresses the issue of missing and fragmented trajectories by associating almost every detection box, regardless of its score. Traditional methods often discard low-confidence detection boxes, leading to significant errors in MOT performance. ByteTrack uses a simple and effective association method, BYTE, which associates high-confidence detection boxes with tracklets and then matches low-confidence detection boxes to these tracklets. This approach helps recover occluded objects and filters out background detections. The method is evaluated on nine state-of-the-art trackers, showing consistent improvements in metrics such as MOTA, IDF1, and HOTA. ByteTrack is also designed as a strong tracker, achieving 80.3 MOTA, 77.3 IDF1, and 63.1 HOTA on the MOT17 dataset with 30 FPS on a single V100 GPU. It outperforms other trackers on various benchmarks, including MOT20, HiEve, and BDD100K. The source code and pre-trained models are available at <https://github.com/ifzhang/ByteTrack>.