RF-DETR-XL and RF-DETR-S do not segment anything

[piotrgrubicki]

1. Create a segmentation project, upload a 720p video (I am not sure if it happens for images as well).
2. Annotate several frames.
3. Train RF-DETR-XL, RF-DETR-S, and Mask-RCNN-ResNet50 using an xpu.
4. Go to Annotate to see how the models perform.

RF-DETR-XL and RF-DETR-S: no object for all the frames.
Mask-RCNN-ResNet50 finds objects on the same frames.