Observation-Centric SORT: Rethinking SORT for Robust Multi-Object Tracking

Kalman filter (KF) based methods for multi-object tracking (MOT) make an assumption that objects move linearly. While this assumption is acceptable for very short periods of occlusion, linear estimates of motion for prolonged time can be highly inaccurate. Moreover, when there is no measurement available to update Kalman filter parameters, the standard convention is to trust the priori state estimations for posteriori update. This leads to the accumulation of errors during a period of occlusion. The error causes significant motion direction variance in practice. In this work, we show that a basic Kalman filter can still obtain state-of-the-art tracking performance if proper care is taken to fix the noise accumulated during occlusion. Instead of relying only on the linear state estimate (i.e., estimation-centric approach), we use object observations (i.e., the measurements by object detector) to compute a virtual trajectory over the occlusion period to fix the error accumulation of filter parameters during the occlusion period. This allows more time steps to correct errors accumulated during occlusion. We name our method Observation-Centric SORT (OC-SORT). It remains Simple, Online, and Real-Time but improves robustness during occlusion and non-linear motion. Given off-the-shelf detections as input, OC-SORT runs at 700+ FPS on a single CPU. It achieves state-of-the-art on multiple datasets, including MOT17, MOT20, KITTI, head tracking, and especially DanceTrack where the object motion is highly non-linear. The code and models are available at \url{https://github.com/noahcao/OC_SORT}.

PDF Abstract CVPR 2023 PDF CVPR 2023 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Multiple Object Tracking CroHD OC-SORT MOTA 67.9 # 2
IDF1 62.9 # 1
HOTA 44.1 # 1
Multi-Object Tracking DanceTrack OC-SORT HOTA 55.1 # 12
AssA 38.0 # 13
MOTA 89.4 # 12
IDF1 54.2 # 12
Multiple Object Tracking KITTI Tracking test OC-SORT MOTA 90.3 # 1
HOTA 76.5 # 1
Multi-Object Tracking MOT17 OC-SORT MOTA 78.0 # 7
IDF1 77.5 # 7
HOTA 63.2 # 6
Multi-Object Tracking MOT20 OC-SORT MOTA 75.9 # 6
IDF1 76.4 # 6
HOTA 62.4 # 6
Multiple Object Tracking SportsMOT OC-SORT HOTA 73.7 # 3
IDF1 74.0 # 4
AssA 61.5 # 3
MOTA 96.5 # 1
DetA 88.5 # 1
Multi-Object Tracking SportsMOT OC-SORT HOTA 73.7 # 3
IDF1 74.0 # 4
AssA 61.5 # 3
MOTA 96.5 # 1
DetA 88.5 # 1


No methods listed for this paper. Add relevant methods here