Yolov5: Effect of resizing images prior to training, Good or Bad?

[redwankarimsony]

❔Effect of resizing images prior to training, Good or Bad?

I want to train yolov5 with a custom dataset that has variable image sizes. I have managed to produce the label files accordingly. I know, yolov5 can resize them at the training time. However, if I resize the images (without keeping the aspect ratio) before training into a fixed shape (640x640), it reduces the training time per epoch but can it hurt the learning performance?

Additional context

[glenn-jocher]

@redwankarimsony recommend you simply follow the custom training tutorial steps, and examine coco128 for what works in a dataset.

[glenn-jocher]

https://github.com/ultralytics/yolov5#tutorials

[Aktcob]

@redwankarimsony Try it and issue

[github-actions[bot]]

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.