Self-supervised Object Tracking with Cycle-consistent Siamese Networks

Self-supervised learning for visual object tracking possesses valuable advantages compared to supervised learning, such as the non-necessity of laborious human annotations and online training. In this work, we exploit an end-to-end Siamese network in a cycle-consistent self-supervised framework for object tracking. Self-supervision can be performed by taking advantage of the cycle consistency in the forward and backward tracking. To better leverage the end-to-end learning of deep networks, we propose to integrate a Siamese region proposal and mask regression network in our tracking framework so that a fast and more accurate tracker can be learned without the annotation of each frame. The experiments on the VOT dataset for visual object tracking and on the DAVIS dataset for video object segmentation propagation show that our method outperforms prior approaches on both tasks.

Authors

• Weihao Yuan 0001
• Michael Yu Wang
• Qifeng Chen 0001

Keywords

• Visualization
• Target tracking
• Annotations
• Object segmentation
• Object tracking
• Proposals
• Task analysis
• Siamese Network
• Learning Network
• Online Training
• Self-supervised Learning
• Region Proposal
• Cycle Consistency
• Tracking Framework
• Visual Task
• Bounding Box
• Object Location
• Target Object
• Optical Flow
• Video Sequences
• Sequence Of Frames
• Image Retrieval
• Consistency Loss
• Cycle Path
• Region Proposal Network
• Image X
• Need For Annotation
• Response Map
• Tracking Network
• Correlation Filter
• Minimum Bounding Box
• Self-supervised Manner
• Background Labeling
• Object Labels
• Subsequent Frames
• Weighting Factor