Semantic Segmentation2016en

ADE20K Scene Parsing Benchmark

20K training, 2K validation images annotated with 150 object categories. Complex scene parsing benchmark.

Samples:22,210
Metrics:mIoU, pixel-accuracy
Paper / WebsiteDownload
Current State of the Art

InternImage-H

Shanghai AI Lab

62.9

mIoU

mIoU Progress Over Time

Showing 2 breakthroughs from Dec 2021 to Nov 2022

56.758.460.161.863.5Dec 2021Nov 2022mIoUDate

Key Milestones

Dec 2021
Mask2Former (Swin-L)

Masked-attention Mask Transformer.

57.3
Source
Nov 2022
InternImage-HCurrent SOTA

Semantic segmentation SOTA.

62.9
+9.8%
Source
Total Improvement
9.8%
Time Span
11m
Breakthroughs
2
Current SOTA
62.9

mIoUPrimary

#ModelScorePaper / CodeDate
1
InternImage-HOpen Source
Shanghai AI Lab		62.9
arxiv-paper
Dec 2025
2
Mask2Former (Swin-L)Open Source
Meta		57.3
arxiv-paper
Dec 2025

Other Semantic Segmentation Datasets

Cityscapes
ADE20K Benchmark - Semantic Segmentation | CodeSOTA