pengtaojiang / segment-anything-clip Goto Github PK

Hi, I see you are planing to add maskclip.

There is another work beyond MaskCLIP and suits SAM fine, in the aspect of CLIP's explainability: https://github.com/xmed-lab/CLIP_Surgery

It's able to guide SAM to achieve text to mask without manual points. Besides, it enhances many open-vocabulary tasks, like segmentation, multi-label classification, multimodal visualization.

This is the jupyter demo:
https://github.com/xmed-lab/CLIP_Surgery/blob/master/demo.ipynb

This is our segmentation results:

This is our heatmap:

To run the code, the labels from clip is needed:
from clip_text import class_names_coco, class_names_ADE_150
Could you help with where to find these class_names files?

File "amp_points.py", line 302, in main
y1, x1, y2, x2 = min(ind[0]), min(ind[1]), max(ind[0]), max(ind[1])
ValueError: min() arg is an empty sequence