Keywords: visual tracking, contrastive learning, hybrid feature, redundant pruning
Abstract: Visual object tracking is a research hotspot in the field of computer vision, and has been widely applied in video surveillance, human-computer interaction, unmanned driving and other fields. At present, the object trackers based on Transformer have good performance, but they still face the challenge of confusing target and background in the feature extraction process. To address this issue, we propose a Hybrid Contrastive Transformer Tracker (HCTrack) in this paper, which combines contrastive learning to improve the ability of distinguishing the target and the background in video. Furthermore, a hybrid feature interaction module is presented to realize multi-level information exchange between the features of template and search regions and capture the target-related semantic information of the search frames comprehensively. Additionally, we design a redundant information pruning module to adaptively eliminate the redundant backgrounds according to the global scene information, thereby reducing the interference of the background to the target feature. HCTrack achieves superior tracking accuracy on the GOT-10k and TrackingNet datasets compared to other state-of-the-art trackers, while maintaining fast inference speed, as the contrastive learning is only implemented during training model.
Primary Area: applications to computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 10035
Loading