How to produce mask for new image, in few-shot inference?

[nguyenthekhoig7]

Hi everyone, after running demo/demo.py to infer in zero-shot successfully, I want to try infer in few-shot mode:

• I downloaded the correct models
• But I stuck at producing the class prototypes(demo/build_protytypes.ipynb), which require images, AND THEIR MASKS.

In the provided demo, images and mask are provided, but I can not find how those masks were made, so that I can make segmentation mask of my new images. I have searched in issues and could not find a similar question.

So my question is:
How do I make segmentation mask for new images? i.e. see my attached image, from left image, how can I make the segmentation on the right?

I appreciate every suggestion, even how small it is, thank you all!

[retazo0018]

Hi @nguyenthekhoig7,

I generated my masks using Segment Anything Model. You can also use SimpleClick.

Best,