| ReCo (DeepLab v2 with ResNet-101 backbone, ImageNet pretrained) | 56.53% | Bootstrapping Semantic Segmentation with Regional Contrast | |
| UniMatch (DeepLab v3+ with ResNet-101) | 73.0 | Revisiting Weak-to-Strong Consistency in Semi-Supervised Semantic Segmentation | |
| GIST and RIST (DeepLabv2 with ResNet101, MSCOCO pre-trained) | 58.70% | The GIST and RIST of Iterative Self-Training for Semi-Supervised Segmentation | - |
| SemiSegContrast (DeepLab v2 with ResNet-101 backbone, MSCOCO pretrained) | 59.4% | Semi-Supervised Semantic Segmentation with Pixel-Level Contrastive Learning from a Class-wise Memory Bank | |
| CutMix (DeepLab v2, ImageNet pre-trained) | 51.2 | Semi-supervised semantic segmentation needs strong, varied perturbations | |
| ClassMix (DeepLab v2 MSCOCO pretrained) | 54.07% | ClassMix: Segmentation-Based Data Augmentation for Semi-Supervised Learning | |
| GuidedMix-Net(DeepLab v2 with ResNet101, ImageNet pretrained) | 56.9% | GuidedMix-Net: Learning to Improve Pseudo Masks Using Labeled Images as Reference | |
| DMT (DeepLab v2 MSCOCO/ImageNet pre-trained) | 54.80% | DMT: Dynamic Mutual Training for Semi-Supervised Learning | |
| SegSDE (MTL decoder with ResNet101, ImageNet pretrained, unlabeled image sequences) | 62.09% | Three Ways to Improve Semantic Segmentation with Self-Supervised Depth Estimation | |
| SemiSegContrast (DeepLab v3+ with ResNet-50 backbone, MSCOCO pretrained) | 64.9% | Semi-Supervised Semantic Segmentation with Pixel-Level Contrastive Learning from a Class-wise Memory Bank | |
| ReCo (DeepLab v3+ with ResNet-101 backbone, ImageNet pretrained) | 60.28% | Bootstrapping Semantic Segmentation with Regional Contrast | |