Metric: Validation mIoU (higher is better)
| # | Model↕ | Validation mIoU▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | SemiOVS (w/ PrevMatch, ResNet-101) | 81.3 | Yes | - | - | Code |
| 2 | Dual Teacher | 81.03 | No | - | - | Code |
| 3 | SemiOVS (w/ UniMatch, ResNet-101) | 81 | Yes | - | - | Code |
| 4 | AllSpark | 80.92 | No | AllSpark: Reborn Labeled Features from Unlabeled... | 2024-03-04 | Code |
| 5 | CorrMatch (Deeplabv3+ with ResNet-101) | 80.9 | No | CorrMatch: Label Propagation via Correlation Mat... | 2023-06-07 | Code |
| 6 | PrevMatch (ResNet-101) | 80.8 | No | Revisiting and Maximizing Temporal Knowledge in ... | 2024-05-31 | Code |
| 7 | PCR (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K) | 80.78 | No | Semi-supervised Semantic Segmentation with Proto... | 2022-10-10 | Code |
| 8 | UniMatch (DeepLab v3+ with ResNet-101) | 80.43 | No | Revisiting Weak-to-Strong Consistency in Semi-Su... | 2022-08-21 | Code |
| 9 | S4MC | 79.85 | No | Semi-Supervised Semantic Segmentation via Margin... | 2023-08-26 | Code |
| 10 | U2PL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K, CutMix) | 79.3 | No | Semi-Supervised Semantic Segmentation Using Unre... | 2022-03-08 | Code |
| 11 | FARCLUSS | 79 | No | FARCLUSS: Fuzzy Adaptive Rebalancing and Contras... | 2025-06-11 | Code |
| 12 | n-CPS (ResNet-101) | 78.97 | No | n-CPS: Generalising Cross Pseudo Supervision to ... | 2021-12-14 | - |
| 13 | PS-MT (DeepLab v3+ with ImageNet-pretrained ResNet-101, single scale inference) | 78.72 | No | Perturbed and Strict Mean Teachers for Semi-supe... | 2021-11-25 | Code |
| 14 | AEL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K) | 78.06 | No | Semi-Supervised Semantic Segmentation via Adapti... | 2021-10-11 | Code |
| 15 | CPCL (DeepLab v3+ with ResNet-101) | 77.16 | No | Conservative-Progressive Collaborative Learning ... | 2022-11-30 | Code |
| 16 | PCT (DeepLab v3+ with ResNet-50 pretrained on ImageNet-1K) | 76.47 | No | - | - | Code |
| 17 | CW-BASS (DeepLab v3+ with ResNet-50) | 76.2 | No | Confidence-Weighted Boundary-Aware Learning for ... | 2025-02-21 | Code |
| 18 | n-CPS (ResNet-50) | 75.85 | No | n-CPS: Generalising Cross Pseudo Supervision to ... | 2021-12-14 | - |
| 19 | CPCL (DeepLab v3+ with ResNet-50) | 74.58 | No | Conservative-Progressive Collaborative Learning ... | 2022-11-30 | Code |
| 20 | ClassMix (DeepLab v2 MSCOCO pretrained) | 72.45 | No | ClassMix: Segmentation-Based Data Augmentation f... | 2020-07-15 | Code |
| 21 | Dense FixMatch (DeepLabv3+ ResNet-101, over-sampling, single pass eval) | 72.04 | No | Dense FixMatch: a simple semi-supervised learnin... | 2022-10-18 | Code |
| 22 | Dense FixMatch (DeepLabv3+ ResNet-50, over-sampling, single pass eval) | 69.02 | No | Dense FixMatch: a simple semi-supervised learnin... | 2022-10-18 | Code |