Image Classification on CUB (Zero-Shot/Generalized Zero-Shot)
81.6Harmonic Mean Top-1 AccZeroDiff
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ZeroDiffBackbone=Res1012024.06 | 81.6 | 80 | 83.2 | — | |
| TFVAEGANVenue=ECCV20, Backbone=Res1012024.06 | 81.1 | 79 | 83.3 | — | |
| f-VAEGANVenue=CVPR19, Backbone=Res101, feature_source=our fine-tune features2024.06 | 80.3 | 77.4 | 83.5 | — | |
| DFCAFlowVenue=TCSVT23, Backbone=Res101, feature_source=our fine-tune features2024.06 | 80 | 77.3 | 82.9 | — | |
| CEGANVenue=CVPR21, Backbone=Res101, feature_source=our fine-tune features2024.06 | 79.9 | 78.9 | 80.9 | — | |
| f-CLSWGANVenue=CVPR18, Backbone=Res1012024.06 | 79.7 | 76.4 | 83.3 | — | |
| SDVAEVenue=ICCV21, Backbone=Res1012024.06 | 77.7 | 81.6 | 74.2 | — | |
| STHS-WGANLearning Setting=Transductive2021.06 | 75.9 | 77.4 | 74.5 | — | |
| VADSVenue=CVPR24, Backbone=ViT, feature_source=other fine-tune features2024.06 | 74.3 | 74.1 | 74.6 | — | |
| CoAR-ZSLVenue=TNNLS'23, Category=Embedding-based2026.04 | 74 | — | — | 79.2 | |
| ReZSLVenue=TIP23, Backbone=Res1012024.06 | 73.8 | 72.8 | 74.8 | — | |
| PSVMAVenue=CVPR23, Backbone=ViT2024.06 | 73.8 | 70.1 | 77.8 | — | |
| ZSLVITVenue=CVPR24, Backbone=ViT2024.06 | 73.6 | 69.4 | 78.2 | — | |
| STHS-S2VLearning Setting=Transductive2021.06 | 72.8 | 71.2 | 74.5 | — | |
| TF-VAEGANLearning Setting=Transductive2021.06 | 71 | 69.9 | 72.1 | — | |
| DVSAVenue=Ours, Category=Embedding-based2026.04 | 70.8 | — | — | 74.9 | |
| GEM-ZSLVenue=CVPR'21, Category=Embedding-based2026.04 | 70.4 | — | — | 77.8 | |
| SABR-TLearning Setting=Transductive2021.06 | 70.3 | 67.2 | 73.7 | — | |
| SDGNLearning Setting=Transductive2021.06 | 70.1 | 69.9 | 70.2 | — | |
| ZS-VATVenue=TNNLS'25, Category=Embedding-based2026.04 | 69.4 | — | — | 66.4 | |
| TranszeroVenue=AAAI'22, Category=Embedding-based2026.04 | 68.8 | — | — | 76.8 | |
| MSDNVenue=CVPR'22, Category=Embedding-based2026.04 | 68.1 | — | — | 76.1 | |
| A-RSRVenue=TNNLS'22, Category=Embedding-based2026.04 | 68 | — | — | 72.1 | |
| D3GZSLVenue=AAAI'24, Category=Generative-based2026.04 | 67.8 | — | — | — | |
| DMLVenue=ACCV24, Backbone=Res1012024.06 | 67.2 | 57.1 | 81.6 | — | |
| APNVenue=NeurIPS'20, Category=Embedding-based2026.04 | 67.2 | — | — | 72 | |
| DAA+SPOTprotocol=Generalized Zero-shot, SPOT_enabled=true2023.04 | 67 | 67.7 | 66.3 | — | |
| BGSNetVenue=TMM'23, Category=Embedding-based2026.04 | 66.7 | — | — | 73.3 | |
| ICCEVenue=CVPR'22, Category=Generative-based2026.04 | 66.4 | — | — | 78.4 | |
| DAAprotocol=Generalized Zero-shot, SPOT_enabled=false2023.04 | 65.8 | 65.5 | 66.1 | — | |
| CE-GZSLVenue=CVPR'21, Category=Generative-based2026.04 | 65.3 | — | — | 77.5 | |
| GMNLearning Setting=Inductive2021.06 | 65 | 60.2 | 70.6 | — | |
| GCM-CFCategory=Generation-based, Protocol=Generalized Zero-Shot Learning (GZSL)2021.03 | 60.3 | 61 | 59.7 | — | |
| CMC-GAN+SPOTprotocol=Generalized Zero-shot, SPOT_enabled=true2023.04 | 59.1 | 66.7 | 53.1 | — | |
| FREE+SPOTprotocol=Generalized Zero-shot, SPOT_enabled=true2023.04 | 58.6 | 62.2 | 55.5 | — | |
| CMC-GANprotocol=Generalized Zero-shot, SPOT_enabled=false2023.04 | 58.2 | 65.1 | 52.6 | — | |
| BSeGNVenue=TNNLS'22, Category=Generative-based2026.04 | 58 | — | — | 65.3 | |
| FREEprotocol=Generalized Zero-shot, SPOT_enabled=false2023.04 | 57.7 | 59.9 | 55.7 | — | |
| CoOp+SHIPVenue=ICCV23, Backbone=ViT, class_splitting=does not follow strict class splitting2024.06 | 57.1 | 55.3 | 58.9 | — | |
| DSPVenue=ICML23, Backbone=Res1012024.06 | 56.9 | 51.4 | 63.8 | — | |
| VFD-ZSLVenue=ICME'24, Category=Generative-based2026.04 | 56.9 | — | — | 62.2 | |
| DAZLELearning Setting=Inductive2021.06 | 56.5 | 53.2 | 60.2 | — | |
| ICISVenue=ICCV23, Backbone=Res1012024.06 | 56.5 | 45.8 | 73.7 | — | |
| TF-VAEGANCategory=Generation-based, Protocol=Generalized Zero-Shot Learning (GZSL)2021.03 | 56 | 50.7 | 62.5 | — | |
| CoOpVenue=IJCV22, Backbone=ViT, class_splitting=does not follow strict class splitting2024.06 | 55.6 | 49.2 | 63.8 | — | |
| HSVAVenue=NeurIPS'21, Category=Generative-based2026.04 | 55.3 | — | — | 62.8 | |
| AREESVenue=TNNLS'22, Category=Generative-based2026.04 | 55.2 | — | — | 65.7 | |
| CLIPVenue=ICML21, Backbone=ViT, class_splitting=does not follow strict class splitting2024.06 | 55 | 55.2 | 54.8 | — | |
| f-VAEGAN+SPOTprotocol=Generalized Zero-shot, SPOT_enabled=true2023.04 | 54.9 | 62.8 | 48.8 | — | |
| NereNET+SPOTprotocol=Generalized Zero-shot, SPOT_enabled=true2023.04 | 54.6 | 58.4 | 51.3 | — | |
| DE-VAELearning Setting=Inductive2021.06 | 54.3 | 52.5 | 56.3 | — | |
| f-VAEGANprotocol=Generalized Zero-shot, SPOT_enabled=false2023.04 | 53.6 | 60.1 | 48.4 | — | |
| NereNETprotocol=Generalized Zero-shot, SPOT_enabled=false2023.04 | 53.6 | 56.5 | 51 | — | |
| Cycle-WGAN+SPOTprotocol=Generalized Zero-shot, SPOT_enabled=true2023.04 | 53.5 | 62.9 | 46.5 | — | |
| DGCNetVenue=TCSVT'23, Category=Generative-based2026.04 | 53.2 | — | — | 71.6 | |
| CADA-VAECategory=Generation-based, Protocol=Generalized Zero-Shot Learning (GZSL)2021.03 | 53 | 50.3 | 56.1 | — | |
| Transferable Contrastive NetworkProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 52.3 | 52.6 | 52 | — | |
| Cycle-WGANprotocol=Generalized Zero-shot, SPOT_enabled=false2023.04 | 52.2 | 60.3 | 46 | — | |
| APNetLearning Setting=Inductive2021.06 | 51.7 | 48.1 | 55.9 | — | |
| LisGANProtocol=Generalized Zero-Shot Learning2019.04 | 51.6 | 46.5 | 57.9 | — | |
| OCD-GZSLLearning Setting=Inductive2021.06 | 51.3 | 44.8 | 59.9 | — | |
| LisGANCategory=Generation-based, Protocol=Generalized Zero-Shot Learning (GZSL)2021.03 | 51.1 | 44.9 | 59.3 | — | |
| WGAN+SPOTprotocol=Generalized Zero-shot, SPOT_enabled=true2023.04 | 51.1 | 60.9 | 44.1 | — | |
| GXELearning Setting=Transductive2021.06 | 51 | 45.4 | 58.1 | — | |
| f-CLSWGANProtocol=Generalized Zero-Shot Learning2019.04 | 49.7 | 43.7 | 57.7 | — | |
| WGANprotocol=Generalized Zero-shot, SPOT_enabled=false2023.04 | 49.7 | 57.7 | 43.7 | — | |
| FGNProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 49.7 | 43.7 | 57.7 | — | |
| f-VAEGAN-D2Learning Setting=Transductive2021.06 | 49.6 | 60.6 | 41.9 | — | |
| VSCLearning Setting=Transductive2021.06 | 47.9 | 33.1 | 86.1 | — | |
| RNetProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 47 | 38.1 | 61.4 | — | |
| RelationNetCategory=Inference-based, Protocol=Generalized Zero-Shot Learning (GZSL)2021.03 | 46.3 | 36.3 | 63.8 | — | |
| PRENLearning Setting=Transductive2021.06 | 43.1 | 35.2 | 55.8 | — | |
| GAZSLProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 41.8 | 31.7 | 61.3 | — | |
| LsrGANLearning Setting=Inductive2021.06 | 40.9 | 44.8 | 37.7 | — | |
| DASCNLearning Setting=Inductive2021.06 | 40.3 | 42.4 | 38.5 | — | |
| f-CLSWGANLearning Setting=Inductive2021.06 | 39.4 | 42.6 | 36.6 | — | |
| DCNProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 38.7 | 28.4 | 60.7 | — | |
| ALECategory=Inference-based, Protocol=Generalized Zero-Shot Learning (GZSL)2021.03 | 34.4 | 23.7 | 62.8 | — | |
| ALEProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 34.4 | 23.7 | 62.8 | — | |
| GAZSLProtocol=Generalized Zero-Shot Learning2019.04 | 34.3 | 23.9 | 60.6 | — | |
| SJEProtocol=Generalized Zero-Shot Learning2019.04 | 33.6 | 23.5 | 59.2 | — | |
| SJEProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 33.6 | 23.5 | 59.2 | — | |
| CDLProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 32.9 | 23.5 | 55.2 | — | |
| DeViSEProtocol=Generalized Zero-Shot Learning2019.04 | 32.8 | 23.8 | 53 | — | |
| DEVISECategory=Inference-based, Protocol=Generalized Zero-Shot Learning (GZSL)2021.03 | 32.8 | 23.8 | 53 | — | |
| DEVISEProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 32.8 | 23.8 | 53 | — | |
| GDANCategory=Generation-based, Protocol=Generalized Zero-Shot Learning (GZSL)2021.03 | 31.6 | 35 | 28.7 | — | |
| GFZSLLearning Setting=Inductive2021.06 | 30.9 | 23.5 | 45.1 | — | |
| LATEMCategory=Inference-based, Protocol=Generalized Zero-Shot Learning (GZSL)2021.03 | 24 | 15.2 | 57.3 | — | |
| LATEMProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 24 | 15.2 | 57.3 | — | |
| EZSLProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 21 | 12.6 | 63.8 | — | |
| SYNCProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 19.8 | 11.5 | 70.9 | — | |
| DEMProtocol=Generalized Zero-Shot Learning2019.04 | 19.4 | 11.1 | 75.1 | — | |
| SSEProtocol=Generalized Zero-Shot Learning2019.04 | 14.4 | 8.5 | 46.9 | — | |
| SSEProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 14.4 | 8.5 | 46.9 | — | |
| SAEProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 13.6 | 7.8 | 54 | — | |
| SYNCProtocol=Generalized Zero-Shot Learning2019.04 | 13.3 | 7.4 | 66.3 | — | |
| CMTProtocol=Generalized Zero-Shot Learning (GZSL), Accuracy Type=average per-class top-1 accuracy2019.08 | 12.6 | 7.2 | 49.8 | — | |
| ALEProtocol=Generalized Zero-Shot Learning2019.04 | 8.7 | 4.6 | 73.7 | — | |
| ESZSLProtocol=Generalized Zero-Shot Learning2019.04 | 4.6 | 2.4 | 70.1 | — |