Use generalized_box_iou_loss

Hi all,
I’m training a model to fit bounding boxes on images. All Bounding boxes are defined by two coordinates (x1,y1,x2,y2). To fit these bounding boxes I first used mse_loss. The loss converges to 0, but the results are still not great enough.
I therefore tried to use generalized_box_iou_loss. Unfortunately this does not work (the loss converges to 1). Any idea what the problem may be?

Best, Jona