Fine-tunning the original model to new tasks with few-shot adaption

[NTUZZH]

Hi thanks for such an amazing work!

I have noticed the authors haven't revealed the training code, if in this way, could anyone please can guide how to modify the EfficientSAM working on the customised dataset? The customised dataset is ready with COCO-seg format! Specially, how to fine-tune the original EfficientSAM onto new tasks with few-shot scenarios.

Thanks with appreciation.

John