scholar.google.com › citations
Oct 22, 2021 · We propose a step-by-step cycle tracking strategy to capture the temporal motion dynamics in an unsupervised manner. We incorporate contrastive ...
We then perform unsupervised tracking by computing the contrastive cycle consistency over the augmented consecutive frames, which simulates the challenging ...
We then perform unsupervised tracking by computing the contrastive cycle consistency over the augmented consecutive frames, which simulates the challenging ...
Abstract: Unsupervised visual tracking has received increasing attention recently. Existing unsupervised visual tracking methods mainly exploit the cycle ...
We then perform unsupervised tracking by computing the contrastive cycle consistency over the augmented consecutive frames, which simulates the challenging ...
May 13, 2021 · In this paper, we propose a novel contrastive learning method which explores the cross-video relation by using cycle-consistency for general ...
Apr 4, 2022 · This provides a unified technique for self-supervised op- tical flow, pose tracking, and video object segmentation. • For optical flow, the ...
Jun 4, 2024 · In this paper, we develop a novel cycle contrastive adversarial framework for unsupervised SID, which mainly consists of cycle contrastive ...
Missing: Tracking. | Show results with:Tracking.
We propose a self-cycle regularization to consider a cycle-consistency property on individual edges in order to prevent converging on noisy matching or trivial.
Contrastive Cycle Consistency Learning for Unsupervised Visual Tracking. J Zhu, C Ma, S Jia, S Xu. Pattern Recognition and Computer Vision: 4th Chinese ...