search

DragonDescentZerotsu / SSM-SAM

Code for WACV 2024, Self-Sampling Meta SAM: Enhancing Few-shot Medical Image Segmentation with Meta-Learning
22 stars 1 forks source link

Can you push code in the models\mmseg\models\sam part? #1

Closed UreMine02 closed 3 months ago

DragonDescentZerotsu commented 3 months ago

I've just uploaded the missing ones, see if it works for you

UreMine02 commented 3 months ago

I saw your new update on github. But I didn't see any changes in models\mmseg\models\sam part (in which include image_encoder.py, prompt_encoder.py,...). I really want to know this part because in current code, I don't see the part showing cross attention between Gaussian mask and query image in the current code as well as you embed Gaussian mask.

DragonDescentZerotsu commented 3 months ago

See now

UreMine02 commented 3 months ago

I saw your new update on github and it works now. I still have one more question: did your project support prompts which are in original SAM?

DragonDescentZerotsu commented 3 months ago

Yes there are prompts but automatically done by randomly choosing a point on the ground truth area (mask area), you can modify the code to make it interactive

UreMine02 commented 3 months ago

Thank you for being so helpful.