Visual Tracker Benchmark
lastmod 2026-11-10created 2019-03-122 edits
Visual Tracker Benchmark (OTB) is a benchmark created to fairly compare the performance of single-object tracking (visual object tracking) algorithms. It provides a collection of video sequences annotated with attributes such as illumination variation, occlusion, and fast motion, along with evaluation metrics such as the success plot and precision plot and evaluation code, so that trackers can be compared under the same conditions. The datasets known as OTB-50 and OTB-100 were widely used, and larger-scale benchmarks such as GOT-10k and LaSOT followed.