Discrimination on Photorealistic discrimination dataset (test)
93.9Realistic AccCLIP
Evaluation Results
| Method | Links | |
|---|---|---|
| CLIPBackbone=ViT-B/16, Pretraining Scale=400M, Zero-shot evaluation=true2024.06 | 93.9 | |
| ImageNetBackbone=ViT-B/16, Pretraining Scale=14.2M, Zero-shot evaluation=true2024.06 | 53 | |
| MAEBackbone=ViT-B/16, Pretraining Scale=1.28M, Zero-shot evaluation=true2024.06 | 52.4 | |
| DINOBackbone=ViT-B/16, Pretraining Scale=1.28M, Zero-shot evaluation=true2024.06 | 50.9 | |
| ViT-B/16 (No pre-training)Backbone=ViT-B/16, Pretraining Scale=None, Zero-shot evaluation=true2024.06 | 49.9 | |
| DINOv2Backbone=ViT-B/16, Pretraining Scale=142M, Zero-shot evaluation=true2024.06 | 48 |