facebookresearch / facebookresearch/sam3
Can SAM3 segment all instance in an image like SAM1?
- Dominant language
- Python
- Stars
- 11.7k
- Forks
- 1.8k
- PR merge metrics
- No merged PRs in 30d
Description
Hi,
I would like to ask whether SAM3 supports segmenting all objects in a single image in a fully automatic manner, similar to the “automatic mask generation” functionality in SAM1.
In SAM1, it was possible to generate masks for all detectable objects without providing prompts. Is there an equivalent feature or recommended workflow in SAM3 for achieving the same result? If not, are there plans to support this capability?
Thank you for your work and clarification.
Best regards,
Contributor guide
Research direction
No source file, test, or entry point is named. Start with the SAM3 inference examples and notebooks described by the repository, then compare their workflow with SAM1's automatic mask generation. Done would require a confirmed supported workflow, or a clearly scoped implementation decision and documented outcome.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- computer-vision, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100