Semantic Segmentation on PASCAL-Context (unseen)
66.7mIoUMaskCLIP+
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MaskCLIP+Pretrain dataset=CLIP400M, Trained on seen classes=true, Trained on unseen classes=false, different set of unseen classes=true2023.01 | 66.7 | — | — | |
| OVSegmentorPretrain dataset=CC4M, Trained on seen classes=false, Trained on unseen classes=false2023.01 | 54.5 | — | — | |
| GroupViTPretrain dataset=CC12M+YFCC15M, Trained on seen classes=false, Trained on unseen classes=false2023.01 | 51.3 | — | — | |
| Fully Sup.Backbone=DeepLabv3+-ResNet1012021.12 | 48.2 | — | — | |
| MaskCLIP+Learning Paradigm=Transductive, Backbone=DeepLabv3+-ResNet101, Pseudo Label Backbone=CLIP-ResNet-502021.12 | 48.1 | 66.7 | 53.3 | |
| GroupViT*Pretrain dataset=CC12M, Trained on seen classes=false, Trained on unseen classes=false2023.01 | 45.3 | — | — | |
| ZS3Net+STLearning Paradigm=Transductive, Self-training=true, Backbone=DeepLabv3+-ResNet1012021.12 | 26 | 20.7 | 23.4 | |
| GroupViT*Pretrain dataset=CC4M, Trained on seen classes=false, Trained on unseen classes=false2023.01 | 24.5 | — | — | |
| CaGNetLearning Paradigm=Inductive, Backbone=DeepLabv3+-ResNet1012021.12 | 23.2 | 18.5 | 21.2 | |
| ZS3NetLearning Paradigm=Inductive, Backbone=DeepLabv3+-ResNet1012021.12 | 19.4 | 12.7 | 15.8 | |
| ViL-SegPretrain dataset=CC12M, Trained on seen classes=false, Trained on unseen classes=false2023.01 | 18.9 | — | — | |
| GaGNetPretrain dataset=None, Trained on seen classes=true, Trained on unseen classes=true2023.01 | 15 | — | — | |
| SIGNPretrain dataset=None, Trained on seen classes=true, Trained on unseen classes=true2023.01 | 14.9 | — | — | |
| ZS3Pretrain dataset=None, Trained on seen classes=true, Trained on unseen classes=true2023.01 | 7.7 | — | — | |
| SPNetPretrain dataset=None, Trained on seen classes=true, Trained on unseen classes=false2023.01 | 4 | — | — |