Train on images without objects

[hoangcuongbk80]

❔Question

Does the model learn from images without objects?

Additional context

80% of my custom dataset is images without objects. Is it make sense to put these images in the training set? Or I just need to train the model on images having objects.

[glenn-jocher]

@hoangcuongbk80 YOLOv5 datasets can contain images with and without objects. If your image has no objects, it does not need a label txt file.

[hoangcuongbk80]

@glenn-jocher

Thanks for your reply.

I mean does the model learn anything from images without objects (empty)? or it just ignores these images.

[kinoute]

@hoangcuongbk80 It helps to reduce the number of False Positives. You can see a full answer here: #2313 (comment)

[hoangcuongbk80]

@kinoute Thank you, it's clear now.

[glenn-jocher]

Loss is computed on all images regardless of contents.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.