论文部分内容阅读
Public benchmark datasets have been widely used to evaluate multi-target tracking algorithms. Ideally, the benchmark datasets should include the video scenes of all scenarios that need to be tested. However, a limited amount of the currently available ben