Learning Effective Multi-modal Trackers via Modality-Sensitive Tuning

ICLR 2025 Conference Submission3465 Authors

24 Sept 2024 (modified: 13 Oct 2024)ICLR 2025 Conference SubmissionEveryoneRevisionsBibTeXCC BY 4.0
Keywords: Multi-modal Tracking; Cross-modal Fine-Tuning
Abstract: This paper tackles the critical issue of constructing multi-modal trackers by effectively adapting the extensive knowledge of pre-trained RGB trackers to auxiliary modalities.To address the challenges, we propose a novel modality sensitivity-aware tuning framework, namely MST, which delicately models the learning process via adaptive tuning of model weights by inherent modality characteristics. Specifically, we first investigate the parameter modality-sensitivity as a criterion for measuring a precise element-wise essentiality for multi-modal adaptation. Then, in the tuning phase, we further leverage such sensitivity to bolster the stability and coherence of multi-modal representations, thereby enhancing generalization capabilities. Extensive experiments showcase the effectiveness of the proposed method, surpassing current state-of-the-art techniques across various multi-modal tracking scenarios and demonstrating remarkable performance even in extreme conditions. The source code will be publicly available.
Supplementary Material: zip
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: 3465
Loading