Model Request: SAM2

[jenriver]

Model description

SAM2 (SAM 2: Segment Anything in Images and Videos)

Open source examples / papers / references

• Paper: https://arxiv.org/abs/2408.00714
• Website: https://www.nvidia.com/en-us/ai/cosmos/
• Model code: https://github.com/facebookresearch/sam2
• Model weights: https://huggingface.co/models?library=sam2&sort=downloads

Why this model?

• New foundation model for promptable visual segmentation that works for both images and videos.
• Huge leap from previous models (SAM) that focused on static images.

[jenriver]

Being addressed in #3.

[jenriver]

Merged in #3.

[jenriver]

Actually, keeping this one open because we want a real example demo, something on par with https://github.com/facebookresearch/sam2/blob/main/notebooks/image_predictor_example.ipynb. 😄