I'm trying to pre-train the Mdetr architecture on just the COCO dataset, however, there is no availability of tokens_positive in the COCO annotations. I was wondering if there's a way to generate tokens_positive for the COCO dataset. Any help is appreciated!!
Thanks.
I'm trying to pre-train the Mdetr architecture on just the COCO dataset, however, there is no availability of tokens_positive in the COCO annotations. I was wondering if there's a way to generate tokens_positive for the COCO dataset. Any help is appreciated!! Thanks.