Unsupervised Deep Tracking

TLDR

The paper proposes an unsupervised visual tracking method that aims to enable robust forward and backward predictions of target objects. The method employs a Siamese correlation filter network trained on large‑scale unlabeled videos, augmented with a multiple‑frame validation scheme and a cost‑sensitive loss to facilitate unsupervised learning. The unsupervised tracker attains accuracy comparable to fully supervised trackers and shows promise for further improvement by leveraging unlabeled or weakly labeled data.

Abstract

We propose an unsupervised visual tracking method in this paper. Different from existing approaches using extensive annotated data for supervised learning, our CNN model is trained on large-scale unlabeled videos in an unsupervised manner. Our motivation is that a robust tracker should be effective in both the forward and backward predictions (i.e., the tracker can forward localize the target object in successive frames and backtrace to its initial position in the first frame). We build our framework on a Siamese correlation filter network, which is trained using unlabeled raw videos. Meanwhile, we propose a multiple-frame validation method and a cost-sensitive loss to facilitate unsupervised learning. Without bells and whistles, the proposed unsupervised tracker achieves the baseline accuracy of fully supervised trackers, which require complete and accurate labels during training. Furthermore, unsupervised framework exhibits a potential in leveraging unlabeled or weakly labeled data to further improve the tracking accuracy.