Hacker News Reader

U-net is a brilliant architecture, and it still seems to beat this model in scaling up the segmentation mask from 256x256 back to the real image. I also don't think unet really benefits from the massive internal feature size given by a the visual transformer used for image encoding.

But I'm impressed by the ability of this model to create a image encoding that is independent of the prompt. I feel like there may be lessons in training approach that can be carried over to unet for a more valuable encoding.

Agreed that Unet has been the most used model for medical imaging for the last 10 years since the initial Unet paper. I think a combination of Llm+VLMs could be a way forward for medical imaging. I tried it out here and it works great. https://chat.vlm.run/c/e062aa6d-41bb-4fc2-b3e4-7e70b45562cf

Same. My use case is ultrasound segmentation. These models struggle, understandably so, with medical imaging.