Hi, congrats on the cool project. Have you tried using this with SAM models? Will it be able to give better segmentation results out of the box or will it require some fine-tuning with scales?
Thank you! I think it can augment the features from SAM vision backbone without any tuning, but if we want to let the decoder to use the new features, we might need to tune the decoder.
Hi, congrats on the cool project. Have you tried using this with SAM models? Will it be able to give better segmentation results out of the box or will it require some fine-tuning with scales?