facebookresearch / facebookresearch/sam2
Is it possible to enable text prompt for sam2 by adding text embedding model?
Open
- Dominant language
- Jupyter Notebook
- Stars
- 19.9k
- Forks
- 2.5k
- PR merge metrics
- No merged PRs in 30d
Description
`Grounding Dino + Sam2` is a popular way to segment object with text, but the detection module `Grounding Dino` may miss some object. So I think if I can add a text embedding model, such as T5, Sam2 maybe can segment by text prompt end-to-end. Is it possible to achieve this?
Contributor guide
Assessment
This issue has not been assessed yet.