how to evaluate MOT result on panda without the groundtruth of test scenes?

GigaVision / PANDA-Toolkit

PANDA dataset tool kit for data visualization, splitting, merging, and result evaluation.

MIT License

93 stars 24 forks source link

how to evaluate MOT result on panda without the groundtruth of test scenes? #5

Closed boringwar closed 3 years ago

boringwar commented 3 years ago

Hi, authors, I'm evaluating my tracking algorithm on PANDA dataset. The 'video_anno' folder contains no ground-truth data for the last 5 scenes. I am wondering how to evaluate the result to compare with your baselines in CVPR 2020 paper "PANDA: A Gigapixel-level Human-centric Video Dataset".

DarkstartsUp commented 3 years ago

Sorry, we are holding the international AI competition (track 2) recently. In order to ensure the fairness of the competition, we have temporarily closed the test set result submission channel. MOT is the second round of the competition. You can submit the results on the competition page after the second round begins (April. 15th). Please refer to the competition page (https://gaiic.tianchi.aliyun.com) for details.