Browse › Computer Vision › Semantic Segmentation › ADE20K val
ADE20K val Benchmark (Semantic Segmentation)
Over time archive 2025-07-28
The chart needs JavaScript; the table below carries every value.
Direction inferred from the metric name, not from the archive: Pixel Accuracy (higher is better). Not inferred (points only, no best-so-far line): mIoU. Points are placed at the row's paper date; 94 of 95 rows carry one.
Results archive 2025-07-28
Archive rows end at the archive snapshot, 2025-07-28: no result published after that date is in this table. Rank is the archive's row order at that snapshot; not re-ranked here. Metric values are the archive's strings. Column headers sort the table in your browser; each row keeps its archive rank.
| Paper | Code | Ran Syntology | Report | ||||||
|---|---|---|---|---|---|---|---|---|---|
| 1 | BEiT-3 | 62.8 | ✓ | Paper | Code | 2022 | linked, not harvested | report | |
| 2 | ViT-CoMer | 62.1 | – | Paper | Code | 2024 | linked, not harvested | report | |
| 3 | EVA | 61.5 | – | Paper | Code | 2022 | 1 of 3 ran · 2 unverified | report | |
| 4 | FD-SwinV2-G | 61.4 | ✓ | Paper | Code | 2022 | 2 of 8 ran · 6 unverified | report | |
| 5 | MaskDINO-SwinL | 60.8 | ✓ | Paper | Code | 2022 | 11 of 13 ran · 2 unverified | report | |
| 6 | OneFormer (InternImage-H, emb_dim=256, multi-scale, 896x896) | 60.8 | – | Paper | Code | 2022 | 0 of 5 ran · 5 unverified | report | |
| 7 | ViT-Adapter-L (Mask2Former, BEiT pretrain) | 60.5 | ✓ | Paper | Code | 2022 | linked, not harvested | report | |
| 8 | SERNet-Former_v2 | 59.35 | ✓ | Paper | Code | 2024 | linked, not harvested | report | |
| 9 | OneFormer (DiNAT-L, multi-scale, 896x896) | 58.6 | – | Paper | Code | 2022 | 0 of 5 ran · 5 unverified | report | |
| 10 | ViT-Adapter-L (UperNet, BEiT pretrain) | 58.4 | ✓ | Paper | Code | 2022 | linked, not harvested | report | |
| 11 | OneFormer (DiNAT-L, multi-scale, 640x640) | 58.4 | – | Paper | Code | 2022 | 0 of 5 ran · 5 unverified | report | |
| 12 | RSSeg-ViT-L(BEiT pretrain) | 58.4 | – | Paper | – | 2022 | no code linked | report | |
| 13 | EoMT (DINOv2-L, single-scale, 512x512) | 58.4 | – | Paper | Code | 2025 | 1 of 2 ran · 1 unverified | report | |
| 14 | OneFormer (Swin-L, multi-scale, 896x896) | 58.3 | – | Paper | Code | 2022 | 0 of 5 ran · 5 unverified | report | |
| 15 | SeMask (SeMask Swin-L FaPN-Mask2Former) | 58.2 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 16 | SeMask (SeMask Swin-L MSFaPN-Mask2Former) | 58.2 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 17 | DiNAT-L (Mask2Former) | 58.1 | – | Paper | Code | 2022 | linked, not harvested | report | |
| 18 | Mask2Former (Swin-L-FaPN, multiscale) | 57.7 | – | Paper | Code | 2021 | 2 of 8 ran · 6 unverified | report | |
| 19 | OneFormer (Swin-L, multi-scale, 640x640) | 57.7 | – | Paper | Code | 2022 | 0 of 5 ran · 5 unverified | report | |
| 20 | SeMask (SeMask Swin-L Mask2Former) | 57.5 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 21 | SenFormer (BEiT-L) | 57.1 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 22 | BEiT-L (ViT+UperNet, ImageNet-22k pretrain) | 57.0 | – | Paper | Code | 2021 | 6 of 11 ran · 5 unverified | report | |
| 23 | SeMask (SeMask Swin-L MSFaPN-Mask2Former, single-scale) | 57.0 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 24 | FaPN (MaskFormer, Swin-L, ImageNet-22k pretrain) | 56.7 | – | Paper | Code | 2021 | 2 of 4 ran · 2 unverified | report | |
| 25 | Mask2Former (Swin-L-FaPN) | 56.4 | – | Paper | Code | 2021 | 2 of 8 ran · 6 unverified | report | |
| 26 | SeMask (SeMask Swin-L MaskFormer) | 56.2 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 27 | CSWin-L (UperNet, ImageNet-22k pretrain) | 55.7 | – | Paper | Code | 2021 | 2 of 6 ran · 4 unverified | report | |
| 28 | MaskFormer (Swin-L, ImageNet-22k pretrain) | 55.6 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 29 | DeiT-L | 55.6 | – | Paper | Code | 2022 | linked, not harvested | report | |
| 30 | Focal-L (UperNet, ImageNet-22k pretrain) | 55.4 | – | Paper | Code | 2021 | 3 of 3 ran · 0 unverified | report | |
| 31 | SegViT ViT-Large | 55.2 | – | Paper | Code | 2022 | linked, not harvested | report | |
| 32 | PatchDiverse + Swin-L (multi-scale test, upernet, ImageNet22k pretrain) | 54.4% | – | Paper | Code | 2021 | linked, not harvested | report | |
| 33 | K-Net | 54.3 | – | Paper | Code | 2021 | 0 of 1 ran · 1 unverified | report | |
| 34 | DEPICT-SA (ViT-L 640x640 multi-scale) | 54.3 | – | Paper | Code | 2024 | linked, not harvested | report | |
| 35 | SenFormer (Swin-L) | 54.2 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 36 | DeiT-B | 54.1 | – | Paper | Code | 2022 | linked, not harvested | report | |
| 37 | MixMIM-L | 53.8 | – | Paper | Code | 2022 | linked, not harvested | report | |
| 38 | Seg-L-Mask/16 (MS, ViT-L) | 53.63 | – | Paper | Code | 2021 | 0 of 7 ran · 7 unverified | report | |
| 39 | Swin-L (UperNet, ImageNet-22k pretrain) | 53.5 | – | Paper | Code | 2021 | 108 of 207 ran · 99 unverified | report | |
| 40 | SeMask (SeMask Swin-L FPN) | 53.5 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 41 | PatchConvNet-L120 (UperNet) | 52.9 | – | Paper | Code | 2021 | 1 of 2 ran · 1 unverified | report | |
| 42 | DEPICT-SA (ViT-L 640x640 single-scale) | 52.9 | – | Paper | Code | 2024 | linked, not harvested | report | |
| 43 | PatchConvNet-B120 (UperNet) | 52.8 | – | Paper | Code | 2021 | 1 of 2 ran · 1 unverified | report | |
| 44 | SegFormer-B5(MS, 87M #Params, ImageNet-1K pretrain) | 51.8 | – | Paper | Code | 2021 | 48 of 86 ran · 38 unverified | report | |
| 45 | Light-Ham (VAN-Huge, 61M, IN-1k, MS) | 51.5 | – | Paper | Code | 2021 | 3 of 3 ran · 0 unverified | report | |
| 46 | CrossFormer (ImageNet1k-pretrain, UPerNet, multi-scale test) | 51.4% | 84.0% | – | Paper | Code | 2021 | 1 of 1 ran · 0 unverified | report |
| 47 | PatchConvNet-B60 (UperNet) | 51.1 | – | Paper | Code | 2021 | 1 of 2 ran · 1 unverified | report | |
| 48 | Light-Ham (VAN-Large, 46M, IN-1k, MS) | 51.0 | – | Paper | Code | 2021 | 3 of 3 ran · 0 unverified | report | |
| 49 | UperNet Shuffle-B | 50.5 | – | Paper | Code | 2021 | 5 of 12 ran · 7 unverified | report | |
| 50 | ELSA-Swin-S | 50.3 | – | Paper | Code | 2021 | 1 of 1 ran · 0 unverified | report | |
| 51 | MixMIM-B | 50.3 | – | Paper | Code | 2022 | linked, not harvested | report | |
| 52 | Twins-SVT-L (UperNet, ImageNet-1k pretrain) | 50.2 | – | Paper | Code | 2021 | 0 of 2 ran · 2 unverified | report | |
| 53 | Seg-B-Mask/16 (MS, ViT-B) | 50.0 | – | Paper | Code | 2021 | 0 of 7 ran · 7 unverified | report | |
| 54 | Swin-B (UperNet, ImageNet-1k pretrain) | 49.7 | – | Paper | Code | 2021 | 108 of 207 ran · 99 unverified | report | |
| 55 | gSwin-S | 49.69 | 83.43 | – | Paper | – | 2022 | no code linked | report |
| 56 | Seg-B/8 (MS, ViT-B) | 49.61 | 83.37 | – | Paper | Code | 2021 | 0 of 7 ran · 7 unverified | report |
| 57 | UperNet Shuffle-S | 49.6 | – | Paper | Code | 2021 | 5 of 12 ran · 7 unverified | report | |
| 58 | Light-Ham (VAN-Base, 27M, IN-1k, MS) | 49.6 | – | Paper | Code | 2021 | 3 of 3 ran · 0 unverified | report | |
| 59 | PatchConvNet-S60 (UperNet) | 49.3 | – | Paper | Code | 2021 | 1 of 2 ran · 1 unverified | report | |
| 60 | DPT-Hybrid | 49.02 | 83.11 | – | Paper | Code | 2021 | 54 of 116 ran · 62 unverified | report |
| 61 | DaViT-S (UperNet) | 48.8 | – | Paper | Code | 2022 | 8 of 15 ran · 7 unverified | report | |
| 62 | ResNeSt-200 | 48.36 | – | Paper | Code | 2020 | 8 of 48 ran · 40 unverified | report | |
| 63 | HRNetV2 + OCR + RMI (PaddleClas pretrained) | 47.98 | – | Paper | Code | 2019 | 4 of 9 ran · 5 unverified | report | |
| 64 | gSwin-T | 47.63 | 82.60 | – | Paper | – | 2022 | no code linked | report |
| 65 | ResNeSt-269 | 47.60 | – | Paper | Code | 2020 | 8 of 48 ran · 40 unverified | report | |
| 66 | UperNet Shuffle-T | 47.6 | – | Paper | Code | 2021 | 5 of 12 ran · 7 unverified | report | |
| 67 | DCNAS | 47.12 | – | Paper | – | 2020 | no code linked | report | |
| 68 | ResNeSt-101 | 46.91 | – | Paper | Code | 2020 | 8 of 48 ran · 40 unverified | report | |
| 69 | Seg-S-Mask/16 (MS, ViT-S) | 46.9 | – | – | – | not matched | report | ||
| 70 | Swin-S (RPE w/ GAB) | 46.41 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 71 | DaViT-B (UperNet) | 46.3 | – | Paper | Code | 2022 | 8 of 15 ran · 7 unverified | report | |
| 72 | CPN(ResNet-101) | 46.27 | – | Paper | Code | 2020 | linked, not harvested | report | |
| 73 | MultiMAE (ViT-B) | 46.2 | – | Paper | Code | 2022 | linked, not harvested | report | |
| 74 | PyConvSegNet-152 | 45.99 | 82.49 | – | Paper | Code | 2020 | linked, not harvested | report |
| 75 | DNL | 45.97 | – | Paper | Code | 2020 | 1 of 3 ran · 2 unverified | report | |
| 76 | CTNet | 45.94 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 77 | ACNet (ResNet-101) | 45.90 | – | Paper | – | 2019 | no code linked | report | |
| 78 | ACNet(ResNet-101) | 45.90 | – | Paper | – | 2019 | no code linked | report | |
| 79 | OCR (HRNetV2-W48) | 45.66 | – | Paper | Code | 2019 | 4 of 9 ran · 5 unverified | report | |
| 80 | EANet (ResNet-101) | 45.33 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 81 | OCR (ResNet-101) | 45.28 | – | Paper | Code | 2019 | 4 of 9 ran · 5 unverified | report | |
| 82 | Asymmetric ALNN | 45.24 | – | Paper | Code | 2019 | 0 of 3 ran · 3 unverified | report | |
| 83 | gSwin-VT | 45.07 | 81.79 | – | Paper | – | 2022 | no code linked | report |
| 84 | LaU-regression-loss | 45.02 | – | Paper | Code | 2019 | linked, not harvested | report | |
| 85 | EncNet (ResNet-101) | 44.65 | – | Paper | Code | 2018 | 1 of 8 ran · 7 unverified | report | |
| 86 | SGR (ResNet-101) | 44.32 | – | Paper | Code | 2018 | linked, not harvested | report | |
| 87 | Auto-DeepLab-L | 43.98 | 81.72 | – | Paper | Code | 2019 | 1 of 4 ran · 3 unverified | report |
| 88 | PSANet (ResNet-101) | 43.77 | – | Paper | Code | 2018 | linked, not harvested | report | |
| 89 | DSSPN (ResNet-101) | 43.68 | – | Paper | – | 2018 | no code linked | report | |
| 90 | PSPNet (ResNet-152) | 43.51% | – | Paper | Code | 2016 | 7 of 29 ran · 22 unverified | report | |
| 91 | PSPNet (ResNet-101) | 43.29% | – | Paper | Code | 2016 | 7 of 29 ran · 22 unverified | report | |
| 92 | HRNetV2 (HRNetV2-W48) | 42.99 | – | Paper | Code | 2019 | 3 of 18 ran · 15 unverified | report | |
| 93 | UperNet (ResNet-101) | 42.66 | – | Paper | Code | 2018 | 9 of 28 ran · 19 unverified | report | |
| 94 | RefineNet (ResNet-152) | 40.70 | – | Paper | Code | 2016 | 1 of 7 ran · 6 unverified | report | |
| 95 | RefineNet (ResNet-101) | 40.20 | – | Paper | Code | 2016 | 1 of 7 ran · 6 unverified | report |
All 95 rows shown. 94 link to a paper page on this site; 6 are marked as using additional training data in the archive. No GitHub stars are tracked; "Code" is the first repository the archive lists for the row. The archive carries no row tags, review links or community-submitted rows for this table; none are shown. archive 2025-07-28
Syntology Ran reads "N of M ran · U unverified": of the M code samples Syntology harvested from repositories linked to that row's paper (joined by arXiv id), N executed on a synthesized input and the other U = M−N are unverified (harvested, no recorded run). It counts code from repositories linked to that row's paper, not this result: the row's number was not reproduced and nothing here is a correctness claim. The other cell texts mean no graph line for the row: "linked, not harvested" (the archive links code, Syntology has not harvested it), "no code linked" (no code link in the archive), "not matched" (the row's paper URL matched no paper on this site). 54 rows have a graph line, from 32 distinct papers; 43 rows (27 papers) have at least one sample that ran. Counting each paper once: Syntology ran 294 of 673 samples; 379 unverified. Separately, 144 of those 673 are pointer-only (licence): the site points at that code rather than redistributing it, a licence property recorded for ran and unverified samples alike; each cell's tooltip carries the row's own pointer-only count. Read from the graph 2026-09-24. Per-sample status is on the paper page.
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections