Why need to freeze for segmentation task after training the normal box model?

facebookresearch / detr

End-to-End Object Detection with Transformers

Apache License 2.0

13.09k stars 2.37k forks source link

Open jeong-tae opened 1 year ago

jeong-tae commented 1 year ago

Just wonder why do we need to freeze the normal box model after the box model is trained. Why do not train both tasks concurrently?

ryanalexmartin commented 6 months ago

Because it will take too much VRAM.