COCO-checkpoint or dataloader?

[quolc]

Congratulations and thank you for sharing great work.

I would like to try a COCO-trained model to segment my images containing objects that aren't seen in PASCAL.
Do you have any plans to release the COCO-trained model mentioned in the paper or a dataloader script allowing us to train a model with COCO by ourselves?

Thanks!