search

MCG-NJU / FCOT

[CVIU] Fully Convolutional Online Tracking
GNU General Public License v3.0
91 stars 12 forks source link

could you give me the OTB results of your comparison method? #6

Open meimeixu520 opened 4 years ago

meimeixu520 commented 4 years ago

Hi,thanks your excellent work, could you give me the OTB results of your comparison method? I look foward your reply! Best wishes!

jcaha commented 4 years ago

I think you need results of all methods displayed in our SOTA comparison of OTB100 dataset. I've put the mat file on Baidu drive https://pan.baidu.com/s/13OJielrB2Rl9kUGXygbzpQ (extract code:s4pw). To use the results, OTB official evaluation toolkit is needed.

meimeixu520 commented 4 years ago

Thank you for sharing,when I use the test files that you Shared, I find the dimensions of the results are different(5,725,4)?

jcaha commented 4 years ago

For ATOM and DiMP method, all tracking results of 5 runs are given. They averaged the 5 runs to report the results. You can refer this page. To reproduce their results of evaluation, we changed the code in genPerfMat.m of the official OTB100 toolkit to average results of all 5 predited bboxes in each frame.