Semantic Segmentation on Cityscapes without background class
27.6mIoUTagAlign
Evaluation Results
| Method | Links | |
|---|---|---|
| TagAlignPost-processing=false, Training Datasets=CC12M, Zero-shot evaluation=true2023.12 | 27.6 | |
| TagAlignPost-processing=true, Training Datasets=CC12M, Zero-shot evaluation=true2023.12 | 27.3 | |
| TCLEvaluation Mode=Zero-shot2022.12 | 24 | |
| TCLPost-processing=false, Training Datasets=CC3&12M, Zero-shot evaluation=true2023.12 | 24 | |
| TCLPost-processing=true, Training Datasets=CC3&12M, Zero-shot evaluation=true2023.12 | 23.1 | |
| ReCoPost-processing=false, Training Datasets=ImageNet1K, Zero-shot evaluation=true2023.12 | 22.8 | |
| CoCuPost-processing=false, Training Datasets=CC3&12M+YFCC, Zero-shot evaluation=true2023.12 | 22.1 | |
| CoCuPost-processing=true, Training Datasets=CC3&12M+YFCC, Zero-shot evaluation=true2023.12 | 21.9 | |
| MaskCLIP (Baseline)Refinement=None2022.12 | 21.6 | |
| ReCoEvaluation Mode=Zero-shot2022.12 | 21.1 | |
| ReCoPost-processing=true, Training Datasets=ImageNet1K, Zero-shot evaluation=true2023.12 | 21.1 | |
| MaskCLIPRefinement=With heuristic refinement2022.12 | 12.6 | |
| Dong et al.Post-processing=true, Training Datasets=LAION-20M, Zero-shot evaluation=true2023.12 | 12.6 | |
| GroupViTPost-processing=false, Training Datasets=CC12M+YFCC, Zero-shot evaluation=true2023.12 | 11.6 | |
| SegCLIPPost-processing=true, Training Datasets=CC12M+COCO, Zero-shot evaluation=true2023.12 | 11.2 | |
| Dong et al.Post-processing=false, Training Datasets=LAION-20M, Zero-shot evaluation=true2023.12 | 11.2 | |
| GroupViT (RedCaps)Training Data=RedCaps + CC12M2022.12 | 11.1 | |
| GroupViT (YFCC)Training Data=YFCC + CC12M2022.12 | 6.9 | |
| GroupViTPost-processing=true, Training Datasets=CC12M+YFCC, Zero-shot evaluation=true2023.12 | 6.9 | |
| OVSegmentorPost-processing=true, Training Datasets=CC12M*, Zero-shot evaluation=true2023.12 | 6.4 |