Dose SAM3 support finetune by a segmentation dataset only with mask annotation?

Can I finetune SAM3 with a segmentation dataset which only having mask annotation (without object detection format annotation like Yolo or coco)?

If yes, could you provide example codes?

Thanks!

Start with the repository README and the example notebooks that cover SAM 3 finetuning and dataset preparation. Determine whether mask-only segmentation data is supported, then document the required annotation format and provide a minimal example if the existing workflow permits it.

Written by the indexing model from the issue text.