Image Classification on ImageNet 1k (val) (Accuracy & Accuracy Gain)
77.64AccuracyMultimodal Concept-based Explanations
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Multimodal Concept-based ExplanationsBackbone=ViT-L/14@336px, Zero-shot=true, Multimodal explanation descriptors=true2024.10 | 77.64 | 2.15 | |
| Multimodal Concept-based ExplanationsBackbone=ViT-L/14, Zero-shot=true, Multimodal explanation descriptors=true2024.10 | 76.74 | 2.05 | |
| CLIPBackbone=ViT-L/14@336px, Zero-shot=true, Multimodal explanation descriptors=false2024.10 | 75.49 | — | |
| CLIPBackbone=ViT-L/14, Zero-shot=true, Multimodal explanation descriptors=false2024.10 | 74.69 | — | |
| Multimodal Concept-based ExplanationsBackbone=RN50x16, Zero-shot=true, Multimodal explanation descriptors=true2024.10 | 72.22 | 3.75 | |
| Multimodal Concept-based ExplanationsBackbone=ViT-B/16, Zero-shot=true, Multimodal explanation descriptors=true2024.10 | 70.28 | 2.35 | |
| CLIPBackbone=RN50x16, Zero-shot=true, Multimodal explanation descriptors=false2024.10 | 68.47 | — | |
| Multimodal Concept-based ExplanationsBackbone=RN50x4, Zero-shot=true, Multimodal explanation descriptors=true2024.10 | 67.93 | 3.57 | |
| CLIPBackbone=ViT-B/16, Zero-shot=true, Multimodal explanation descriptors=false2024.10 | 67.93 | — | |
| Multimodal Concept-based ExplanationsBackbone=ViT-B/32, Zero-shot=true, Multimodal explanation descriptors=true2024.10 | 65.58 | 2.3 | |
| CLIPBackbone=RN50x4, Zero-shot=true, Multimodal explanation descriptors=false2024.10 | 64.36 | — | |
| Multimodal Concept-based ExplanationsBackbone=RN101, Zero-shot=true, Multimodal explanation descriptors=true2024.10 | 64.14 | 3.46 | |
| CLIPBackbone=ViT-B/32, Zero-shot=true, Multimodal explanation descriptors=false2024.10 | 63.28 | — | |
| Multimodal Concept-based ExplanationsBackbone=RN50, Zero-shot=true, Multimodal explanation descriptors=true2024.10 | 61.85 | 2.31 | |
| CLIPBackbone=RN101, Zero-shot=true, Multimodal explanation descriptors=false2024.10 | 60.68 | — | |
| CLIPBackbone=RN50, Zero-shot=true, Multimodal explanation descriptors=false2024.10 | 59.54 | — |