nodes/kijai-ComfyUI-segment-anything-2

T

kijai 39209164b9 possible device fixes

2024-08-02 14:29:50 +03:00

.github/workflows

chore(publish): Add Github Action for Publishing to Comfy Registry

2024-07-31 22:59:18 +00:00

support individual objects and bboxes for single image detection

2024-08-02 01:37:57 +03:00

fixes

2024-08-02 12:39:19 +03:00

add video mode, rest of the models

2024-07-30 19:37:42 +03:00

__init__.py

initial commit

2024-07-30 16:37:43 +03:00

.gitattributes

Initial commit

2024-07-30 16:36:40 +03:00

.gitignore

initial commit

2024-07-30 16:37:43 +03:00

LICENSE

Create LICENSE

2024-07-31 02:35:42 +03:00

load_model.py

cleanup

2024-08-01 03:16:54 +03:00

nodes.py

possible device fixes

2024-08-02 14:29:50 +03:00

pyproject.toml

registry

2024-08-01 03:12:00 +03:00

readme.md

Update readme.md

2024-07-31 02:38:01 +03:00

requirements.txt

initial commit

2024-07-30 16:37:43 +03:00

readme.md

WORK IN PROGRESS

https://github.com/user-attachments/assets/c1efb595-0fb1-4ae7-b4fa-2def08eda0a8

For testing only currently.

Functional, but needs better coordinate selector.

For now mask postprocessing is disabled due to it needing cuda extension compilation. We can use other nodes for this purpose anyway, so might leave it that way, we'll see.

Models are automatically downloade from https://huggingface.co/Kijai/sam2-safetensors/tree/main

to ComfyUI/models/sam2

Original repo:

https://github.com/facebookresearch/segment-anything-2