Enhancing Multi-Object Tracking with Siamese Network-based Appearance Search

다국어 초록 (Multilingual Abstract)

Multi-target tracking has made significant progress in recent years as a key player in the field of computer vision, but remains a challenging problem due to target similarity and complexity. In recent years, deep learning has pushed the field forward, and detection-based tracking methods use an end-to-end strategy that unifies target detection and trajectory modeling in a neural network framework, however, little use is made of appearance information.
In this paper, we propose a Siamese structure-based approach that introduces an appearance search branch, aiming to enhance the system's ability to model the utilization of target appearance information. The method is validated on the basis of the FairMOT model, which generates a heat map reflecting the results of the target appearance search by means of feature vectors with multiple time dimensions and the Siamese module. The results of the detection branch and the appearance search branch are fused to form a final multi-target tracking system through post-processing and matching. Experiments demonstrate that the method achieves significant performance improvements over existing methods. This research provides a new perspective on the multi-target tracking problem, enhances the modeling and use of target appearance information through the appearance search branch, and provides an effective tool for system performance improvement in complex scenarios.

참고문헌 (Reference)

1 J. Redmon, "You Only Look Once: Unified, Real-Time Object Detection" 779-788, 2016

2 X. Zhou, "Tracking Objects as Points" 12349 : 474-490, 2020

3 Z. Wang, "Towards Real-Time Multi-Object Tracking" 12356 : 107-122, 2020

4 N. Wojke, "Simple online and realtime tracking with a deep association metric" 3645-3649, 2017

5 A. Bewley, "Simple online and realtime tracking" 3464-3468, 2016

6 R. Tao, "Siamese Instance Search for Tracking" 1420-1429, 2016

7 J. Pang, "Quasi-Dense Similarity Learning for Multiple Object Tracking" 164-173, 2021

8 B.-S. Hua, "Pointwise Convolutional Neural Networks" 984-993, 2018

9 J. Cao, "Observation-Centric SORT: Rethinking SORT for Robust Multi-Object Tracking" 9686-9696, 2023

10 F. Zeng, "MOTR: End-to-End Multiple-Object Tracking with Transformer" 13687 : 659-675, 2022

11 J. Luiten, "HOTA: A Higher Order Metric for Evaluating Multi-object Tracking" 129 (129): 548-578, 2021

12 L. Bertinetto, "Fully-Convolutional Siamese Networks for Object Tracking" 9914 : 850-865, 2016

13 S. Ren, "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks" 39 (39): 1137-1149, 2017

14 Y. Zhang, "FairMOT : On the Fairness of Detection and Re-identification in Multiple Object Tracking" 129 (129): 3069-3087, 2021

15 X. Chen, "Exploring Simple Siamese Representation Learning" 15745-15753, 2021

16 P. Sun, "Dancetrack: Multi-Object Tracking in Uniform Appearance and Diverse Motion" 20993-21002, 2022

17 Y. Zhang, "ByteTrack : Multi-object Tracking by Associating Every Detection Box" 13682 : 1-21, 2022