I'm currently testing the MedSAM2 solution and trying to use the model without prior knowledge of the segmented mask during inference. I'm considering using a grid prompt instead, which would allow me to perform inference on unknown scans. In this case, is it sufficient to use the grid only during inference, or should it also be used during training? If I checked correctly, such options don't seem to be available in MedSAM at the moment, correct?
Yes, it was not supported because this function is relatively not useful in medical images. I recommend using the official SAM2 repo for the grid prompt.
Hi,
I'm currently testing the MedSAM2 solution and trying to use the model without prior knowledge of the segmented mask during inference. I'm considering using a grid prompt instead, which would allow me to perform inference on unknown scans. In this case, is it sufficient to use the grid only during inference, or should it also be used during training? If I checked correctly, such options don't seem to be available in MedSAM at the moment, correct?
Thanks!