Codesota · Benchmark · SA-1BHome/Leaderboards/SA-1B
Unknown

SA-1B.

Segment Anything benchmark with 1B+ masks across 11M images

Paper ↗Leaderboard ↓
§ 01 · SOTA history

Year over year.

§ 02 · Leaderboard

Results by metric.

Only 4 models on this benchmark
Help build the community leaderboard — submit your model results.
Found a wrong score or missing run?
Use row edits to send a sourced correction into moderation.
Add / edit result ↗Report issue ↗

Miou

Miou is the reported evaluation metric for SA-1B. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for Miouverifiedpapervendorcommunityunverified

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

RankModelTrustScoreYearLinksFix
01SAM 2 (Hiera-L)
SAM 2 (Segment Anything Model 2) Hiera-L. Zero-shot image segmentation on 23 diverse benchmarks. mIoU. Meta AI, 2024. Table 5.
verified62.22024Source ↗Looks wrong?
02SAM (ViT-H)
SAM ViT-H. Zero-shot mask quality on 23 diverse datasets (SA-23). mIoU. Meta AI, ICCV 2023. Table 5.
verified58.12023Source ↗Looks wrong?
03FastSAM
FastSAM (YOLOv8-based). Evaluated on SA-23 benchmark. mIoU comparable to SAM but ~50x faster. Table 1.
verified57.12023Source ↗Looks wrong?
04EfficientSAM
EfficientSAM (SAMI pretraining). Evaluated on 12 zero-shot segmentation benchmarks. mIoU. Meta AI, CVPR 2024. Table 1.
verified55.52023Source ↗Looks wrong?
§ 04 · Submit a result

Add to the leaderboard.

← Back to Leaderboards