Skip to content

Latest commit

 

History

History
 
 

deepsort

Simple online and realtime tracking with a deep association metric

Abstract

Simple Online and Realtime Tracking (SORT) is a pragmatic approach to multiple object tracking with a focus on simple, effective algorithms. In this paper, we integrate appearance information to improve the performance of SORT. Due to this extension we are able to track objects through longer periods of occlusions, effectively reducing the number of identity switches. In spirit of the original framework we place much of the computational complexity into an offline pre-training stage where we learn a deep association metric on a largescale person re-identification dataset. During online application, we establish measurement-to-track associations using nearest neighbor queries in visual appearance space. Experimental evaluation shows that our extensions reduce the number of identity switches by 45%, achieving overall competitive performance at high frame rates.

Citation

@inproceedings{bewley2016simple,
  title={Simple online and realtime tracking},
  author={Bewley, Alex and Ge, Zongyuan and Ott, Lionel and Ramos, Fabio and Upcroft, Ben},
  booktitle={2016 IEEE International Conference on Image Processing (ICIP)},
  pages={3464--3468},
  year={2016},
  organization={IEEE}
}
@inproceedings{wojke2017simple,
  title={Simple online and realtime tracking with a deep association metric},
  author={Wojke, Nicolai and Bewley, Alex and Paulus, Dietrich},
  booktitle={2017 IEEE international conference on image processing (ICIP)},
  pages={3645--3649},
  year={2017},
  organization={IEEE}
}

Results and models on MOT17

We implement SORT and DeepSORT with independent detector and ReID models. To train a model by yourself, you need to train a detector following here and also train a ReID model following here. The configs in this folder are basically for inference.

Currently we do not support training ReID models. We directly use the ReID model from Tracktor. These missed features will be supported in the future.

Method Detector ReID Train Set Test Set Public Inf time (fps) MOTA IDF1 FP FN IDSw. Config Download
SORT R50-FasterRCNN-FPN - half-train half-val Y 28.3 46.0 46.6 289 82451 4581 config detector
SORT R50-FasterRCNN-FPN - half-train half-val N 18.6 62.0 57.8 15171 40437 5841 config detector
DeepSORT R50-FasterRCNN-FPN R50 half-train half-val Y 20.4 48.1 60.8 283 82445 1199 config detector reid
DeepSORT R50-FasterRCNN-FPN R50 half-train half-val N 13.8 63.8 69.6 15060 40326 3183 config detector reid

Note: When running demo_mot.py, we suggest you use the config containing private, since private means the MOT method doesn't need external detections.