Train with negative sample (No table in image)

[namirinz]

Hi I would like to know how to train table detection or does it support if there are not any table in the image and XML of that image wouldn't have object. I tried to train but it give me an error which declare that there are no label (object) in XML file.

[bsmock]

Yes, this is possible when training DETR, but we did not write the data loader to handle empty training samples.

I believe it would be just a few lines of code to make the change. If you make the change yourself, feel free to submit a pull request!

[mspondon]

Hey Guys, Any update on this ? My concern is the above detection model detect a non tabular region as a table. I want to finetune the model with negative sample.