The Multiple Object Tracking Benchmark https://motchallenge.net/
高速跟踪: 当检测精度较高,视频帧率较高时,跟踪问题就会变得很简单,主要是多阈值目标检测和 判断前后帧的重合率 High-Speed Tracking-by-Detection Without Using Image Information Advanced Video and Signal Based Surveillance (AVSS), 2017 14th IEEE International Conference on Assessing Post-Detection Filters for a Generic Pedestrian Detector in a Tracking-By-Detection Scheme https://github.com/bochinski/iou-tracker/
密集人头检测 End-to-end people detection in crowded scenes CVPR2016 Code: https://github.com/Russell91/reinspect Evaluation of ReInspect: http://nbviewer.jupyter.org/github/Russell91/ReInspect/blob/master/evaluation_reinspect.ipynb
人体躯干检测 DPM Object detection with discriminatively trained part-based models http://www.rossgirshick.info/latent/
行人检测算法 Aggregated Channel Feature (ACF) detector Fast feature pyramids for object detection. PAMI, 36(8):1532–1545, 2014
the winner of the MOT17 challenge A Novel Multi-Detector Fusion Framework for Multi-Object Tracking 针对多目标检测跟踪问题,这里从检测和跟踪两个方面进行了改进,检测上采用多个检测器融合来提升检测效果(人头检测+躯干检测),跟踪上设计新的 data association models: graph labeling problem
多目标跟踪 Benchmark Multiple Object Tracking Benchmark https://motchallenge.net/ https://motchallenge.net/results/MOT17Det/
Tracking the Trackers: An Analysis of the State of the Art in Multiple Object Tracking 本文针对多目标跟踪问题,给出了两个基准测试数据库: MOT15, MOT16,给出了 50个跟踪算法在这两个数据集上的跟踪效果。