OCTrack: Benchmarking the Open-Corpus Multi-Object Tracking
This work addresses the need for standardized evaluation in open-corpus multi-object tracking, which is an incremental advancement for researchers in computer vision and tracking.
The paper tackles the problem of open-corpus multi-object tracking by introducing OCTrackB, a large-scale benchmark for evaluating methods that localize, associate, and recognize objects across seen and unseen classes without category prompts, reporting results from state-of-the-art methods to demonstrate its utility.
We study a novel yet practical problem of open-corpus multi-object tracking (OCMOT), which extends the MOT into localizing, associating, and recognizing generic-category objects of both seen (base) and unseen (novel) classes, but without the category text list as prompt. To study this problem, the top priority is to build a benchmark. In this work, we build OCTrackB, a large-scale and comprehensive benchmark, to provide a standard evaluation platform for the OCMOT problem. Compared to previous datasets, OCTrackB has more abundant and balanced base/novel classes and the corresponding samples for evaluation with less bias. We also propose a new multi-granularity recognition metric to better evaluate the generative object recognition in OCMOT. By conducting the extensive benchmark evaluation, we report and analyze the results of various state-of-the-art methods, which demonstrate the rationale of OCMOT, as well as the usefulness and advantages of OCTrackB.