Revisiting the details when evaluating a visual tracker
This work addresses the problem of fair and accurate tracker evaluation for researchers and practitioners in computer vision, but it is incremental as it builds on existing benchmarks.
The paper revisits evaluation details for single object visual trackers using the OTB benchmark, proposing a simpler and more accurate method for comparison, and finds no absolute winner among trackers, emphasizing the need for detailed analysis to select suitable ones for specific use cases.
Visual tracking algorithms are naturally adopted in various applications, there have been several benchmarks and many tracking algorithms, more expected to appear in the future. In this report, I focus on single object tracking and revisit the details of tracker evaluation based on widely used OTB\cite{otb} benchmark by introducing a simpler, accurate, and extensible method for tracker evaluation and comparison. Experimental results suggest that there may not be an absolute winner among tracking algorithms. We have to perform detailed analysis to select suitable trackers for use cases.