Siamese Visual Object Tracking: A Survey

Object tracking belongs to active research areas in computer vision. We are interested matching-based trackers exploiting deep machine learning known as Siamese trackers. Their powerful capabilities stem from similarity learning. This paradigm is promising due its inherent balance between performance and efficiency, so of this type suitable for real-time generic object tracking. There an upsurge interest the lack available specialized surveys category. In survey, we aim identify elaborate on most significant challenges face. Our goal answer what design decisions authors made problems they attempted solve first place. thus perform in-depth analysis core principles which operate with a discussion incentives behind them. Besides, provide up-to-date qualitative quantitative comparison prominent established benchmarks. Among other things, discuss current trends developing survey could help absorb details about underlying