A field guide to the Segment Anything explosion

What it does This repository is a living companion to the first comprehensive survey on Meta’s Segment Anything Model (SAM). It catalogs the deluge of research spawned by SAM—seminal papers, yearly follow-ups, and open-source projects—across images, video, 3D, audio, and medical imaging. Think of it as a community-maintained appendix to an academic field guide.

The interesting bit The maintainers treat GitHub as a rolling bibliography rather than a codebase, accepting pull requests to keep pace with a foundation-model ecosystem that now includes SAM 2, SAM 3, SAM 3D, and even SAM Audio. The repo also corrals domain-specific offshoots like medical-image and remote-sensing surveys that would otherwise scatter across arXiv.

Key highlights

• Curated paper lists from 2023 through 2026, covering SAM through SAM 3.1.
• Aggregated domain-specific surveys (medical imaging, video, remote sensing, efficient variants).
• Community-driven updates via PRs to a shared spreadsheet.
• Companion to the original survey paper analyzing SAM’s strengths and limitations across software, real-world, and complex scenes.
• Tracks major awards: ICCV 2023 and ICLR 2025 Best Paper Honorable Mentions for SAM and SAM 2.

Caveats

• The highlights timeline lists future release dates (e.g., 2026.03.27 for SAM 3.1), which may confuse strictly linear readers.
• This is pure curation—expect bibliographies and links, not code or benchmarks.

Verdict Academic researchers and PhD students navigating the SAM literature should bookmark this. Developers hunting ready-to-use segmentation models should keep scrolling.