Joint detection and tracking in videos with identification features

作者：

Highlights：

• We formulate detection, tracking and re-id into an end-to-end joint model.

• Our model is on par with state-of-the-art detection and tracking techniques.

• Our model outperforms state-of-the-art tracking methods at low frame rates.

• We rank 3rd in UA DETRAC’18 tracking challenge and 1st among online trackers.

• We achieve state-of-the-art performance on the MOT dataset.

摘要

•We formulate detection, tracking and re-id into an end-to-end joint model.•Our model is on par with state-of-the-art detection and tracking techniques.•Our model outperforms state-of-the-art tracking methods at low frame rates.•We rank 3rd in UA DETRAC’18 tracking challenge and 1st among online trackers.•We achieve state-of-the-art performance on the MOT dataset.

论文关键词：Detection,Multi-object tracking,Re-identification,Online,Tracking by detection

论文评审过程：Received 20 June 2019, Revised 6 December 2019, Accepted 7 May 2020, Available online 18 May 2020, Version of Record 23 May 2020.

论文官网地址：https://doi.org/10.1016/j.imavis.2020.103932