IndexError，objects_id和instance_id-1不相等

[fiht86]

Hello,

I'm encountering some issues while trying to reproduce your code. In a specific example, I found that the pixel values in an image exceed the instance_id - 1 value. Although there are pixel values reaching 38, the objects_id only goes up to 37. This discrepancy leads to an IndexError during training. Upon further investigation, I discovered that a total of 217 images exhibit similar errors. Could you please confirm if the dataset labels are accurate?

Your assistance in resolving this matter would be greatly appreciated.

Thank you!

[Jingkang50]

Thank you for your interest in our work!
Could you help use this notebook to investigate the annotation? I am out of computational quota to check the problem recently, so if you could help check it out, please do so and can email me for a zoom chat if the problem persists.

Thanks!

[fiht86]

Thank you! I will try it!

[fiht86]

Thank you for your interest in our work! Could you help use this notebook to investigate the annotation? I am out of computational quota to check the problem recently, so if you could help check it out, please do so and can email me for a zoom chat if the problem persists.

Thanks!

Hello:

I tried the method you provided and had encountered a few issues:

1、Missing Categories for Instance IDs:
I have identified four videos where certain instance IDs are missing associated categories.
This is one of them.

2、Label Mismatch:
Additionally, I have noticed discrepancies between some of the provided labels and the actual content in the images.
This is one of them.

Could you please advise on how best to address these issues? Thank you for your attention to this matter. I look forward to your response.

[QiueY514]

I also encountered this issue.

[zef1611]

Hi I am also encountered this issue, have anyone know how to fix this?

[Jingkang50]

Could you email me to organize a zoom chat to share more information?

[zef1611]

Could you email me to organize a zoom chat to share more information?

Thank you for your kindness and responsibility. I’ll reach out to you once I’ve compiled a list of potential error data from the training process. I really appreciate your time and look forward to discussing everything further.

Best,
Huy

[leaozhun]

Hello,have u solved this problem?I also meet it.