Semantic Segmentation on Pascal VOC (Seen/Unseen Evaluation)
93.6Seen mIoUPOMP
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| POMPmode=open-vocabulary2023.04 | 93.6 | — | — | — | — | 76.8 | — | — | — | 84.4 | |
| MVP-SEG+Learning Setting=Transductive, ST (Self-Training)=true2023.04 | 89 | — | — | — | — | 87.4 | — | — | 88.6 | 88.2 | |
| MaskCLIP+Learning Setting=Transductive, ST (Self-Training)=false2023.04 | 88.1 | — | — | — | — | 86.1 | — | — | 87.4 | 86.7 | |
| ZegFormerLearning Setting=Inductive, ST (Self-Training)=false2023.04 | 86.4 | — | — | — | — | 63.6 | — | — | — | 73.3 | |
| ZegFormermode=open-vocabulary2023.04 | 86.4 | — | — | — | — | 63.6 | — | — | — | 73.3 | |
| FCNsetting=zero-shot, Self-Training=false2021.12 | 85.5 | — | — | — | — | 36 | — | — | — | 50.7 | |
| ZSSegsetting=zero-shot, Self-Training=false2021.12 | 83.5 | — | — | — | — | 72.5 | — | — | — | 77.5 | |
| SIGNLearning Setting=Transductive, ST (Self-Training)=true2023.04 | 83.5 | — | — | — | — | 41.3 | — | — | — | 55.3 | |
| ZSSegmode=open-vocabulary2023.04 | 83.5 | — | — | — | — | 72.5 | — | — | — | 77.5 | |
| SIGNBackbone=Deeplab-v2, Embeddings=word2vec/fasttext, Self-training=false, Annealed Self-training=true2021.08 | 83.49 | — | — | — | — | 41.29 | — | — | — | 55.26 | |
| STRICTsetting=zero-shot, Self-Training=false2021.12 | 82.7 | — | — | — | — | 35.6 | — | — | — | 49.8 | |
| STRICTLearning Setting=Transductive, ST (Self-Training)=false2023.04 | 82.7 | — | — | — | — | 35.6 | — | — | 70.9 | 49.8 | |
| ZSSeg+STsetting=zero-shot, Self-Training=true2021.12 | 79.2 | — | — | — | — | 78.1 | — | — | — | 79.3 | |
| SIGNBackbone=Deeplab-v2, Embeddings=word2vec/fasttext, Self-training=true, Annealed Self-training=false2021.08 | 78.62 | — | — | — | — | 33.12 | — | — | — | 46.61 | |
| CaGNet+STsetting=zero-shot, Self-Training=true2021.12 | 78.6 | — | — | — | — | 30.3 | — | — | — | 43.7 | |
| CaGNetLearning Setting=Transductive, ST (Self-Training)=true2023.04 | 78.6 | — | — | — | — | 30.3 | — | — | 65.8 | 43.7 | |
| CaGNetZero-shot protocol=true, Backbone=DeepLabv2, Self-training (ST)=true, Calibration=false2020.08 | 78.59 | 87.04 | 83.9 | 58.55 | 50.71 | 30.31 | 81.64 | 75.6 | 65.77 | 43.66 | |
| CaGNetBackbone=Deeplab-v2, Embeddings=word2vec/fasttext, Self-training=true, Annealed Self-training=false2021.08 | 78.59 | — | — | — | — | 30.31 | — | — | — | 43.66 | |
| CaGNetZero-shot protocol=true, Backbone=DeepLabv2, Self-training (ST)=false, Calibration=false2020.08 | 78.4 | 89.5 | 88.68 | 42.97 | 39.4 | 26.59 | 80.68 | 76.36 | 65.45 | 39.72 | |
| CaGNetsetting=zero-shot, Self-Training=false2021.12 | 78.4 | — | — | — | — | 25.6 | — | — | — | 39.7 | |
| CaGNetLearning Setting=Inductive, ST (Self-Training)=false2023.04 | 78.4 | — | — | — | — | 26.6 | — | — | 65.5 | 39.7 | |
| CaGNetmode=open-vocabulary2023.04 | 78.4 | — | — | — | — | 25.6 | — | — | — | 39.7 | |
| CaGNetBackbone=Deeplab-v2, Embeddings=word2vec/fasttext, Self-training=false, Annealed Self-training=false2021.08 | 78.4 | — | — | — | — | 26.59 | — | — | — | 39.72 | |
| ZS3NetZero-shot protocol=true, Backbone=DeepLabv2, Self-training (ST)=true, Calibration=false2020.08 | 78.02 | 91.89 | 85.69 | 34.07 | 26.37 | 21.15 | 80.95 | 73.82 | 63.02 | 33.28 | |
| ZS3Backbone=Deeplab-v2, Embeddings=word2vec/fasttext, Self-training=true, Annealed Self-training=false2021.08 | 78.02 | — | — | — | — | 21.15 | — | — | — | 33.28 | |
| SPNetZero-shot protocol=true, Backbone=DeepLabv2, Self-training (ST)=false, Calibration=true2020.08 | 78 | 88.77 | 87.91 | 29.55 | 23.87 | 15.63 | 77.55 | 71.88 | 63.15 | 26.1 | |
| ZS5setting=zero-shot, Self-Training=false2021.12 | 78 | — | — | — | — | 21.2 | — | — | — | 33.3 | |
| SPNetLearning Setting=Inductive, ST (Self-Training)=false2023.04 | 78 | — | — | — | — | 15.6 | — | — | 63.2 | 26.1 | |
| ZS3NetLearning Setting=Transductive, ST (Self-Training)=true2023.04 | 78 | — | — | — | — | 21.2 | — | — | 63 | 33.3 | |
| SPNetBackbone=Deeplab-v2, Embeddings=word2vec/fasttext, Self-training=false, Annealed Self-training=false2021.08 | 78 | — | — | — | — | 15.63 | — | — | — | 26.1 | |
| SPNet+STsetting=zero-shot, Self-Training=true2021.12 | 77.8 | — | — | — | — | 25.8 | — | — | — | 38.8 | |
| SPNetLearning Setting=Transductive, ST (Self-Training)=true2023.04 | 77.8 | — | — | — | — | 25.8 | — | — | 64.8 | 38.8 | |
| JointLearning Setting=Inductive, ST (Self-Training)=false2023.04 | 77.7 | — | — | — | — | — | — | — | — | 45.9 | |
| ZS3NetZero-shot protocol=true, Backbone=DeepLabv2, Self-training (ST)=false, Calibration=false2020.08 | 77.3 | 92.96 | 87.72 | 21.47 | 15.8 | 17.65 | 79.41 | 73.49 | 61.64 | 28.74 | |
| ZS3setting=zero-shot, Self-Training=false2021.12 | 77.3 | — | — | — | — | 17.7 | — | — | — | 28.7 | |
| ZS3NetLearning Setting=Inductive, ST (Self-Training)=false2023.04 | 77.3 | — | — | — | — | 17.7 | — | — | 61.6 | 28.7 | |
| ZS3mode=open-vocabulary2023.04 | 77.3 | — | — | — | — | 17.7 | — | — | — | 28.7 | |
| ZS3Backbone=Deeplab-v2, Embeddings=word2vec/fasttext, Self-training=false, Annealed Self-training=false2021.08 | 77.3 | — | — | — | — | 17.65 | — | — | — | 28.74 | |
| SPNetZero-shot protocol=true, Backbone=DeepLabv2, Self-training (ST)=false, Calibration=false2020.08 | 75.83 | 94.82 | 94.58 | 0.07 | 0.01 | 0.01 | 76.85 | 70.93 | 56.87 | 0.02 | |
| ZS5NetK (Number of unseen classes)=2, Self-training=true, Zero-shot=true2019.06 | 75.7 | 94.3 | 85.2 | 89.5 | 89.9 | 75.8 | 94.2 | 85.9 | 75.8 | 75.7 | |
| SIGNLearning Setting=Inductive, ST (Self-Training)=false2023.04 | 75.4 | — | — | — | — | 28.9 | — | — | — | 41.7 | |
| SIGNBackbone=Deeplab-v2, Embeddings=word2vec/fasttext, Self-training=false, Annealed Self-training=false2021.08 | 75.4 | — | — | — | — | 28.86 | — | — | — | 41.74 | |
| ZS5NetK (Number of unseen classes)=4, Self-training=true, Zero-shot=true2019.06 | 74 | 93.9 | 84.8 | 57.5 | 62.9 | 53 | 92.4 | 80.9 | 69.8 | 61.8 | |
| SPNetsetting=zero-shot, Self-Training=false2021.12 | 73.3 | — | — | — | — | 15 | — | — | — | 25.1 | |
| SPNetmode=open-vocabulary2023.04 | 73.3 | — | — | — | — | 15 | — | — | — | 21.8 | |
| ZS5NetK (Number of unseen classes)=10, Self-training=true, Zero-shot=true2019.06 | 72.3 | 90.1 | 83.9 | 57.8 | 48 | 34.5 | 86.8 | 66.9 | 54.4 | 46.7 | |
| ZS5NetK (Number of unseen classes)=6, Self-training=true, Zero-shot=true2019.06 | 71.2 | 93.8 | 82 | 68.3 | 61.2 | 53.1 | 92.1 | 75.8 | 66.1 | 60.8 | |
| ZS5NetK (Number of unseen classes)=8, Self-training=true, Zero-shot=true2019.06 | 68.3 | 92.6 | 77.8 | 68.2 | 62 | 50 | 90.2 | 71.9 | 61.3 | 57.7 | |
| CLIPSegMechanism=CLIP text prompts + decoder, Notes=Requires prompts per class2026.04 | 58 | — | — | — | — | 20 | — | — | — | — | |
| CaGNetinference_mode=inductive2021.12 | — | 89.5 | 88.7 | 43 | 39.4 | — | 80.7 | 76.4 | — | — | |
| CaGNetinference_mode=transductive2021.12 | — | 87 | 83.9 | 58.6 | 50.7 | — | 81.6 | 75.6 | — | — | |
| Fully SupST (Self-Training)=false2023.04 | — | — | — | — | — | — | — | — | 88.2 | — | |
| Fully Sup.supervision=fully supervised2021.12 | — | — | — | — | — | — | 94 | 93.4 | — | — | |
| MaskCLIP+inference_mode=transductive2021.12 | — | 94.6 | 93.7 | 91.4 | 92.6 | — | 94 | 93.4 | — | — | |
| SPNetinference_mode=inductive2021.12 | — | 94.8 | 94.6 | 0 | 0 | — | 76.9 | 70.9 | — | — | |
| SPNet-Cinference_mode=inductive2021.12 | — | 88.8 | 87.9 | 29.6 | 23.9 | — | 77.6 | 71.9 | — | — | |
| ZS3Netinference_mode=inductive2021.12 | — | 93 | 87.7 | 21.5 | 15.8 | — | 79.4 | 73.5 | — | — | |
| ZS3Netinference_mode=transductive2021.12 | — | 91.9 | 85.7 | 34.1 | 26.4 | — | 81 | 73.8 | — | — |