vot evaluate problem!!

[Zhangyong-Tang]

The method is always stopped at ' request = self._trax.wait() ' . Wether it is correct that this step takes a long time or not?

[Zhangyong-Tang]

The problem is '''unable to connect to tracker'''

[lukacu]

Please provide some more information on how did you integrate your tracker, perhaps share tracker config and wrapper code. From the current information we cannot really say much more than that there is an error in either wrapper or config.

[Zhangyong-Tang]

manager = vot.VOTManager(tracker, "mask")
manager.run()

vot = 0.6.3

nothing in the toolkit changed

[Zhangyong-Tang]

[label] label = label protocol = traxpython command = test_vots2023 paths = /data/Disk_A/zhangyong/VOT2023/master

[Zhangyong-Tang]

[Zhangyong-Tang]

i have tried on both the Ubantu and Windows systems, on two machines, but both stopped at this line.

[lukacu]

Did you run the CLI test? Can you show me the output that it produced to see what are the Trax messages? Also, did you try to run one of the demo trackers?

[Zhangyong-Tang]

I have passed the CLI TEST. Everything seems well.

Actually, it seems that 'vot evaluate name' haven't got in touch with my tracker, it just break during the connection process.

[lukacu]

Yes, this is indeed strange, the test should be more or less the same thing. Can you try perhaps also running test with a real sequence (you can specify a sequence path and point it to one of the real sequences in VOTS2023 dataset). So far I have no idea what could be the problem.

[davidyang180]

Yes, this is indeed strange, the test should be more or less the same thing. Can you try perhaps also running test with a real sequence (you can specify a sequence path and point it to one of the real sequences in VOTS2023 dataset). So far I have no idea what could be the problem.

Strangely, I had the same problem, but I was able to pass request = self._trax.wait() normally a few days ago, and when I ran the evaluation again today, I got this error.

[davidyang180]

I have passed the CLI TEST. Everything seems well.

Actually, it seems that 'vot evaluate name' haven't got in touch with my tracker, it just break during the connection process.

Hi! May I ask where this test script is mentioned? I don't seem to see it, I'll try this too.

[Zhangyong-Tang]

@davidyang180 My script is 'vot test ncc' Just follow the offical NCC test, but i only pass the test with the label in rectangle manner, failed in the mask manner

[davidyang180]

@Zhangyong-Tang I tried it in mask manner just now, and the error was reported as Tracker does not exist. rectangle manner label? Does the VOTS2023 dataset have rectangle label? Or did you test it with last year's VOT dataset.

[Zhangyong-Tang]

Use the python_ncc rather than the python_ncc_mask to pass the rectangle test.

[lukacu]

I will check all the examples in the integration repository, but note that some of them are not suitable for multi-object tracking.

[lukacu]

We are fixing tutorials since we have noticed some inconsistent information (a problem when different people work on a tight deadline). I have now pushed improved examples that were verified to work with VOTS2023 dataset (at least locally), multiobject examples are clearly named this way.

[JornyWan]

Hello, I also want to ask how to convert the gt file in the four sequences of dev dataset into bbox. I saw that you said this gt file was masked. If I want to use the data of these four seqs to test my tracker, how should I export this gt file (for example, the first cat has 1542 images, the gt file has 1542 lines, and every three lines have the last two actions 0, how should I convert the contents of the first line into bbox)?

Thank you so much!

[Zhangyong-Tang]

I think you can find this in one closed issue 76

[JornyWan]

I think you can find this in one closed issue 76

Hi, i want to ask if you have used these four sequences to test the tracker? Are you using io files to decode and then testing?

Thanks.

[lukacu]

I am closing the topic now since it is not really an issue, discussion belongs to the forum.