Python VOT toolkit - beta testing

[Luka Čehovin Zajc]

As promised more than a year ago, we are now releasing a new generation of VOT toolkit for public beta testing. Due to a shift in the community the new toolkit is written in Python programming language which should be more aligned with the current dominant preferences in the research community. Our goal is that the toolkit is used for the upcoming VOT2020 challenge so help is welcome. In the following weeks we will accept suggestions and error reports and try to incorporate them in a beta version that will be available when the challenge starts.

A guide on setting up the toolkit and tracker is available here.

A few remarks:

• Trackers integrated with the old toolkit (with TraX) should work out-of-the-box with the new one. But we strongly advise to update the TraX wrappers from the Github integration repository.
• The new toolkit pre-release does not support the previous VOT benchmarks yet – use the old toolkit for that. This release is just to test for the VOT2020.
• The new toolkit does not yet implement all the performance measures – this is just to test the integration.

[Mohamed Hamed]

Thank you!

I think the Tracking Community has been waiting on this for so long:) to avoid all issues that came with the old versions.

[Zh Hipeng]

Hi, I met a problem when I config this new toolkit.

```

ModuleNotFoundError: No module named 'python_ncc'

```

Could you kindly help me to fix it?

On Thursday, March 5, 2020 at 7:57:39 AM UTC-8, Luka Čehovin Zajc wrote:

[Michael Teresi]

try cloning https://github.com/votchallenge/integration.git

and updating your trackers.ini file to have the entry:

`paths =<path to your cloned integration.git project>/integration/python`

[laisi...@gmail.com]

So far(October 20)， does python version toolkit still just support vot2020 evaluation? Or you have upgrate them?

[luka.c...@gmail.com]

Well, in theory, it supports also other evaluation measures as well as older datasets, but we are still debugging some inconsistencies in scores in comparison to Matlab toolkit. Until we at least know their cause we cannot officially support them because people would start opening tickets. And since I am working on another project at the moment, the progress is slow. But if anyone is willing to test older datasets and help us with debugging, this would speed up the progress.

[laisi...@gmail.com]

Ok, if I encounter some error info, I will feedback to you!