WARNING /content/datasets/Insectivisor-35/train/images/1006_jpg.rf.2c4fd3b77488ed1c97e6829bcb00d1eb.jpg: ignoring corrupt image/label: non-normalized or out of bounds coordinates
It is clear after checking the files as well that the label files contain bounding boxes with values above 1 which is unnatural for the YOLOv8 export option that I chose. I have also tried exporting to YOLOv5 format to see if this could solve it with no change at all.
What I have done prior to this and might be related is that I added new labels and 2 new classes in the project. Regardless, I have excluded the new classes in the preprocessing.
This was tested on dataset versions: v35, v36 and v37 all with the same results. It should be noted that v35 was created before I added new labels and annotations, while the other 2 after.
I don’t know what changed since then, but the warnings seem to be gone. I can still try and generate a version without Auto-Orient if that will be helpful to you. (v41)
That’s great to hear! I’m glad your issue got solved. If there’s anything you changed or did that caused your issue to get solved, listing it here might help other people experiencing the same issue.