효율적인 검출 기반 추적 메커니즘을 위한 자가 학습 기반 검출기 스케줄러
A Self-Supervised Detector Scheduler for Efficient Tracking-by-Detection Mechanism
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
본 논문에서는 실시간 고성능 다중 객체 추적을 수행하기 위해 최적의 TBD (Tracking-by-detection) 메커니즘을 결정할 수 있는 Detector Scheduler를 제안한다. Detector Scheduler는 서로 다른 프레임 간의 특징량 차이를 측정하는 것으로 검출기 실행 여부를 결정하여 전체 추적 속도를 향상한다. 하지만, Detector Scheduler의 학습에 필요한 GT (Ground Truth) 생성이 어렵기 때문에 Detector Scheduler를추적 결과만을 통해 학습 가능한 자가 학습 방법을 제안한다. 제안된 자가 학습 방법은 프레임 간의객체 카디널리티와 객체 외형 특징량의 비유사도가 커질 때 검출기를 실행할 수 있도록 의사 레이블을 생성하고 제안된 손실함수를 통해 Detector Scheduler를 학습한다.
Abstract
In this paper, we propose the Detector Scheduler which determines the best tracking-by-detection (TBD) mechanism to perform real-time high-accurate multi-object tracking (MOT). The Detector Scheduler determines whether to run a detector by measuring the dissimilarity of features between different frames. Furthermore, we propose a self-supervision method to learn the Detector Scheduler with tracking results since it is difficult to generate ground truth (GT) for learning the Detector Scheduler. Our proposed self-supervision method generates pseudo labels on whether to run a detector when the dissimilarity of the object cardinality or appearance between frames increases. To this end, we propose the Detector Scheduling Loss to learn the Detector Scheduler. As a result, our proposed method achieves real-time high-accurate multi-object tracking by boosting the overall tracking speed while keeping the tracking accuracy at most.
참고문헌 (30)
- Wang, Zhongdao, "Towards real-time multi-object tracking", European Conference on Computer Vision, pp.767~770 (2020)
- Wojke, Nicolai, "Simple online and realtime tracking with a deep association metric", 2017 IEEE international conference on image processing (ICIP), pp.107~122 (2020)
- Lu, Zhichao, "Retinatrack: Online single stage joint detection and tracking", Proceedings of the IEEE, 2020, 14668~14678
- Zhang, Yifu, "Fairmot: On the fairness of detection and re-identification in multiple object tracking", International Journal of Computer Vision, 129(11), pp.3069~3087 (2021)[DOI]
- Zhu, Xizhou, "Towards high performance video object detection", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.7210~7218 (2018)
- Luo, Hao, "Detect or track: Towards cost-effective video object detection", tracking, 2019, pp.33(01), 8803~8810
- Chen, Kai, "Optimizing video object detection via a scale-time lattice", Proceedings of the IEEE conference on computer vision and pattern recognition, pp.7814~7823 (2018)
- Yu, Fengwei, "Poi: Multiple object tracking with high performance detection and appearance feature", European Conference on Computer Vision, pp.36~42 (2016)
- Feichtenhofer, Christoph, "Detect to track and track to detect", Proceedings of the IEEE international conference on computer vision, pp.3038~3046 (2017)
- Zhu, Xizhou, "Deep feature flow for video recognition", Proceedings of the IEEE conference on computer vision and pattern recognition, pp.2349~2358 (2017)
- Yao, Chun-Han, "Video object detection via object-level temporal aggregation", European conference on computer vision, pp.160~177 (2020)
- Chen, Ting, "Self-supervised gans via auxiliary rotation loss", Proceedings of the IEEE, 2019, 12154~12163
- Noroozi, Mehdi, "Boosting self-supervised learning via knowledge transfer", Proceedings of the IEEE conference on computer vision and pattern recognition, pp.9359~9367 (2018)
- Wang, Qiang, "Multiple object tracking with correlation learning", Proceedings of the IEEE, 2021, 3876~3886
- Dosovitskiy, Alexey, "Flownet: Learning optical flow with convolutional networks", Proceedings of the IEEE international conference on computer vision, pp.2758~2766 (2015)
- Karthik, Shyamgopal, "Simple unsupervised multi-object tracking", ~ (2020)
- Kalman, Rudolph Emil, "A new approach to linear filtering and prediction problems", Journal of Fluids Engineering, 82, pp.35~45 (1960)[DOI]
- Kuhn, Harold W, "The Hungarian method for the assignment problem", Naval research logistics quarterly, 2(1-2), pp.83~97 (1955)[DOI]
- Milan, Anton, "MOT16: A benchmark for multi-object tracking", ~ (2016)
- Bernardin, Keni, "Evaluating multiple object tracking performance: the clear mot metrics", EURASIP Journal on Image and Video Processing, pp.1~10 (2008)
- Pang, Jiangmiao, "Quasi-dense similarity learning for multiple object tracking", Proceedings of the IEEE, 2021, 164~173
- Peng, Jinlong, "Chained-tracker: Chaining paired attentive regression results for end-to-end joint multiple-object detection and tracking", European conference on computer vision, pp.145~161 (2020)
- Babaee, "A dual cnn–rnn for multiple people tracking", Neurocomputing, 368, pp.69~83 (2019)[DOI]
- Lin, Xufeng, "On the detection-to-track association for online multi-object tracking", Pattern Recognition Letters, 146, pp.200~207 (2021)[DOI]
- Wu, Jialian, "Track to detect and segment: An online multi-object tracker", Proceedings of the IEEE, 2021, 12352~12361
- Pang, Bo, "Tubetk: Adopting tubes to track multi-object in a one-step training model", Proceedings of the IEEE, 2020, 6308~6318
- Wang, Yongxin, "Joint object detection and multi-object tracking with graph neural networks", 2021 IEEE International Conference on Robotics and Automation (ICRA), pp.13708~13715 (2021)
- Zhou, Xingyi, "Tracking objects as points", European Conference on Computer Vision, pp.474~490 (2020)
- Zheng, Linyu, "Improving multiple object tracking with single object tracking", Proceedings of the IEEE, 2021, 2453~2462
- Tokmakov, Pavel, "Learning to track with object permanence", Proceedings of the IEEE, 2021, 10860~10869