Publication record · 18.cifr/2023.kirillov.segment-anything

Segment Anything

v1.0.0

Alexander Kirillov (Meta AI Research), Eric Mintun (Meta AI Research), Nikhila Ravi (Meta AI Research), Hanzi Mao (Meta AI Research), Chloe Rolland (Meta AI Research), Laura Gustafson (Meta AI Research), Tete Xiao (Meta AI Research), Spencer Whitehead (Meta AI Research), Alexander C. Berg (Meta AI Research), Wan-Yen Lo (Meta AI Research), Piotr Dollar (Meta AI Research), Ross Girshick (Meta AI Research)

RAI18.cifr/2023.kirillov.segment-anything

ICCV 2023· 2023· doi:10.1109/ICCV51070.2023.00371

We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive -- often competitive with or even superior to prior fully supervised results.

image segmentationfoundation modelpromptable segmentationzero-shot transfervision transformer

✦ Research context

What this agent contributes to the literature.

Problem solved

Before SAM, segmentation required task-specific models and training data for each new domain or task type. SAM provides a single foundation model transferable zero-shot to new image distributions and segmentation tasks, eliminating the need for task-specific retraining in many practical settings.

Novelty

SAM introduces a promptable segmentation paradigm enabling a single model to handle diverse segmentation tasks zero-shot via flexible geometric prompts. The SA-1B dataset of 1B+ masks was collected using SAM itself in a data engine loop — an unprecedented scale. The architecture separates a heavy image encoder from a lightweight prompt-conditioned mask decoder, enabling real-time interactive use.

Related research

Computing related research...

Canvas contract1-in / 1-out · unpacked into image_input, model_params legacy ports

Sample data

Loading sample data...

Total calls

This month

Citations

Last called

—

Image digest

sha256:fe8aa27b21d73ee5c4f0c326b521a9258fa0f3ff8747d4d9cbd402c382d0990f

Invoke command

python main.py

Inputs

input:application/json

Outputs

output:application/json

Citation

Loading DOI…

Invoke

CPU compute only

How to get GPU access: Your university, lab, or company can become a CIFR institutional member. Members get GPU-accelerated runs for all their researchers. Contact us

Recent invocations(0)

No invocations yet — be the first to call this agent.