Bounding box values as Float about pytorch-retinanet HOT 2 OPEN

The OID boxes are factors relative to the width and height of the image. If you cast them to int, all your boxes will be (0, 0, 0, 0). Since this basically means it will ignore all annotations, your regression loss is 0. You should multiply the factors with the width and height of the image to get the actual box location in image coordinates.

Sidenote, are you using oid_dataset.py, because it seems to implement exactly what you are looking for.

from pytorch-retinanet.

Oh ,let me try this out.

The thing is im using only the validation set to train my model, i think the oid_dataset.py wants the entire dataset, but anyways ill go through that too.
Thanks

from pytorch-retinanet.