do evaluation on OTB and VOT about pymdnet

[jimmy-dq]

@zhangyujia0223 , hello, do you evaluate the model which is provided by the author on the OTB-100 dataset? I test it and find that it is lower than the matlab version.

[zhangyujia0223]

@Handworker I don't know how to do evaluation with python. I am beginner. Could you give me some examples? Thank you very much! I want to discuss with you.

[jimmy-dq]

@zhangyujia0223 Maybe you can write a script to test it on the OTB dataset (e.g., python run_tracker -s xx -d -f ), you only need to obtain all the sequence names and specify the name xx in the above.

[zhangyujia0223]

@Handworker yeah, I have done that. But I want to do the evaluation on OTB and VOT (one pass evaluation--OPE). Especially I want to draw the precision plot and success plot. I think there are example codes to do that. But I didn't find it. Could you do that?

[jimmy-dq]

@zhangyujia0223 I only evaluate it on OTB with the matlab implementation. If you want to do that, firstly convert the results (i.e., json file) to the .mat file, and then you can follow the benchmark steps to obtain the OPE evaluation. Hope that it is helpful for you.

[zhangyujia0223]

@Handworker Thanks very much. I'll try.

[zhangyujia0223]

@Handworker I have gotten the .mat file of MDNet, how can I do OPE evaluation just on MDNet? Is there specifical function to do it? The benchmark is for too many trackers.