July 2017, Honolulu, Hawaii, USA
In Conjunction with the Conference on Computer Vision and Pattern Recognition - CVPR 2017
MOT17 Results

We had 2 submissions for MOT17Det and 11 submissions for MOT17. The results can be found here and here

We congratulate the winners and thank all participants.

General Challenge Rules

  • Deadline for submission is on CMT is April 12th, 2017!
  • Deadline for challenge submission on motchallenge.net is July 15th, 2017
  • All presented papers must submit the results of their method to the
    MOT Challenge.
  • The rules of the Challenge must be respected.
  • We will award the best performing method with a prize at the end of the workshop.
  • All submissions will be handled electronically through the submission website.

Challenge 1: Low-density scenarios

Welcome to the CVPR 2017 Detection and Tracking in low-density scenarios Challenge!

This challenge follows the benchmarking of detection and tracking algorithms as set in PETS 2016.  The task is to detect and/or track objects in all frames from video sequences corresponding to this category and report detected/tracked objects at each frame in the MOT Challenge. Authors will be addressing the video sequences contained in the PETS datasets where the density of targets to track is low.  While this task has been addressed traditionally by the community, it still remains an open problem. Not only well known challenges on illumination and occlusion still perturb tracking algorithms but also tracking on different application domains bring specific challenges. Tracking algorithms in the maritime domain, for instance, have to deal with water wakes and waves to accurately detect and track an object, even as a single target. PETS datasets will allow authors to evaluate detection and tracking algorithms on two domains: (1) pedestrian and (2) maritime applications where the density of targets to track is low but the conditions are challenging.

Challenge 2: Detection in Crowded Scenes

Welcome to the CVPR 2017 Detection in Crowded Scenes Challenge!

We are proud to announce the first edition of the Detection Challenge on the challenging Multiple Object Tracking Benchmark, MOT16!
While the benchmark has been previously focusing on tracking with a fixed set of detections, this time we want to see how far state-of-the-art can go from the detection side. We encourage authors to submit their detection results on the challenging MOT16 sequences! These detections will become public for everyone to use for tracking purposes.

Challenge 3: Tracking in Crowded Scenes

Welcome to the CVPR 2017 Tracking in Crowded Scenes Challenge!

Following the success of the 1st and 2nd editions of the BMTT Tracking Challenge at WACV 2015 and ECCV 2016, we bring the challenge to CVPR!
The MOT16 benchmark has still plenty to offer! We challenge you to beat the state of the art in multiple object tracking!

This workshop is part of

Participating institutions