Clarification needed on datasets used for UltraSam.pth pretrained model

[priteeharmale13-hash]

I’m using the UltraSam.pth checkpoint from this repository and would like to clarify the dataset composition:

The README mentions US-43d (43 datasets, 20 clinical applications) covering many organs/regions. GitHub

Could you please confirm whether the model-training included ALL the listed ultrasound domains (abdominal organs, kidney, thyroid, cardiac, nerve, etc.), or whether it was restricted to a subset (e.g., only abdomen/kidney)?

If a subset was used, could you provide a breakdown of which domains and how many samples from each were included?

Thanks for the clarification & for making this great work open!

[Chop1]

Hi, the checkpoint shared in the repo is trained (SAM style, prompted segmentation) with all the dataset mentionned.

[priteeharmale13-hash]

I checked prediction using ultrasam.pth model on samples images but not getting segmentation.