qdtrack

Quasi-Dense Similarity Learning for Multiple Object Tracking

Abstract

Similarity learning has been recognized as a crucial step for object tracking. However, existing multiple object tracking methods only use sparse ground truth matching as the training objective, while ignoring the majority of the informative regions on the images. In this paper, we present Quasi-Dense Similarity Learning, which densely samples hundreds of region proposals on a pair of images for contrastive learning. We can directly combine this similarity learning with existing detection methods to build Quasi-Dense Tracking (QDTrack) without turning to displacementregression or motion priors. We also find that the resulting distinctive feature space admits a simple nearest neighbor search at the inference time. Despite its simplicity, QD-Track outperforms all existing methods on MOT, BDD100K, Waymo, and TAO tracking benchmarks. It achieves 68.7 MOTA at 20.3 FPS on MOT17 without using external training data. Compared to methods with similar detectors, it boosts almost 10 points of MOTA and significantly decreases the number of ID switches on BDD100K and Waymo datasets.

Citation

@inproceedings{pang2021quasi,
  title={Quasi-dense similarity learning for multiple object tracking},
  author={Pang, Jiangmiao and Qiu, Linlu and Li, Xia and Chen, Haofeng and Li, Qi and Darrell, Trevor and Yu, Fisher},
  booktitle={Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},
  pages={164--173},
  year={2021}
}

Results and models on MOT17

Method	Detector	Train Set	Test Set	Public	Inf time (fps)	HOTA	MOTA	IDF1	FP	FN	IDSw.	Config	Download
QDTrack	Faster R-CNN	half-train	half-val	N	-	57.1	68.2	68.5	8373	42939	1071	config	model | log
QDTrack	Faster R-CNN	CrowdHuman + half-train	half-val	N	-	59.1	71.7	71.6	6072	38733	867	config	model | log

Results and models on LVIS dataset

Method	Detector	Train Set	Test Set	Inf time (fps)	AP	AP50	AP75	AP_S	AP_M	AP_L	Config	Download
QDTrack	Faster R-CNN	LVISv0.5+COCO2017	TAO validation	-	17.2	28.6	17.7	5.3	13.0	22.1	config	model | log

Results and models on TAO dataset

Note: If you want to achieve a track AP of 11.0 on the TAO dataset, you need to do pre-training on LVIS dataset.

a. Pre-train the QDTrack on LVISv0.5+COCO2017 training set.

The pre-trained checkpoint is given above(model).

b. Save the model to ckpts/tao/**.pth, and modify the configs for TAO accordingly(set load_from to your ckpt path).

We observe around 0.5 track AP fluctuations in performance, and provide the best model.

Method	Detector	Train Set	Test Set	Inf time (fps)	Track AP(50:75)	Track AP50	Track AP75	Config	Download
QDTrack	Faster R-CNN	TAO train	TAO validation	-	11.0	15.8	6.1	config	model | log