Unsupervised Deep Representation Learning for Real-Time Tracking
This work addresses the need for more efficient and scalable training in visual tracking, particularly for applications requiring real-time performance, though it is incremental as it builds on existing Siamese correlation filter networks.
The authors tackled the problem of reducing manual annotation workload in visual tracking by proposing an unsupervised learning method that learns from unlabeled videos using bidirectional tracking consistency, achieving baseline accuracy comparable to fully supervised trackers with real-time speed.
The advancement of visual tracking has continuously been brought by deep learning models. Typically, supervised learning is employed to train these models with expensive labeled data. In order to reduce the workload of manual annotations and learn to track arbitrary objects, we propose an unsupervised learning method for visual tracking. The motivation of our unsupervised learning is that a robust tracker should be effective in bidirectional tracking. Specifically, the tracker is able to forward localize a target object in successive frames and backtrace to its initial position in the first frame. Based on such a motivation, in the training process, we measure the consistency between forward and backward trajectories to learn a robust tracker from scratch merely using unlabeled videos. We build our framework on a Siamese correlation filter network, and propose a multi-frame validation scheme and a cost-sensitive loss to facilitate unsupervised learning. Without bells and whistles, the proposed unsupervised tracker achieves the baseline accuracy as classic fully supervised trackers while achieving a real-time speed. Furthermore, our unsupervised framework exhibits a potential in leveraging more unlabeled or weakly labeled data to further improve the tracking accuracy.