search

votchallenge / toolkit

The official VOT Challenge evaluation and analysis toolkit
http://www.votchallenge.net/
GNU General Public License v3.0
164 stars 46 forks source link

The context of the output .bin file? #94

Closed Zhangyong-Tang closed 1 year ago

Zhangyong-Tang commented 1 year ago

Hello! With your kind help, i have passeed the mask ncc example, thank you very much.

But some other prooblems appear.

  1. What's the context in the .bin file? I use the matlab code 'A = fread(bin_file,'ubit8');' to read the .bin file. But i am confused about the context. It's not a rectangle or a mask. the size of A is 1496x1, one number per line. image

  2. The analysis I think the format of .bin might be encoded, but can be decoded by the server. So i try to analysis it. But the report seems wrong. image Only the version and time is reported. At first i think may be it's special for the NCC example. So I tried to intergate a simple tracker, but report shows no numerical results again. The data reported to VOT is under the form of List(np.array, np.array,...,number of objects). I think it's consistent with your new NCC example, ncc_multitarget_mask.py.

So would you mind explain these two questions?

Zhangyong-Tang commented 1 year ago

Wether the confidence is a necessary input for the report or not?

lukacu commented 1 year ago

The bin file is a new, more compact format for trajectories. It is binary, but is not obfuscated in any way. The reason for empty analysis report os that groundtruth is not public except for the first frame.

There is no meed to report confidence, it will be ignored. But you have to report of an object is not visible with an empty region (we are upading tutorials on this matter).

Zhangyong-Tang commented 1 year ago

You mean we will not get the quantative results anymore, do i understand correctly?

lukacu commented 1 year ago

Read participation instructions, there is an evaluation server that you can use.

Zhangyong-Tang commented 1 year ago

Thank you very much for patience, I am sorry I havenot notived the website.