Image-Text Matching on MSCOCO HN
49.07I-to-T ScoreOpenCLIP (ViT-B-32)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OpenCLIP (ViT-B-32)Paradigm=Zero-shot, Approach=Discriminative, Backbone=ViT-B-322024.03 | 49.07 | 47.45 | 48.26 | |
| CLIP (ViT-B-32)Paradigm=Zero-shot, Approach=Discriminative, Backbone=ViT-B-322024.03 | 47.63 | 42.82 | 45.23 | |
| DPT*Paradigm=Fine-tuning, Approach=Generative, Priority discriminative task selection=true2024.03 | 43.12 | 35.25 | 39.18 | |
| DPT (Stage 1)Paradigm=Fine-tuning, Approach=Generative2024.03 | 42.29 | 34.75 | 38.52 | |
| DPTParadigm=Fine-tuning, Approach=Generative2024.03 | 42.07 | 34.97 | 38.52 | |
| Diffusion Classifier +Paradigm=Zero-shot, Approach=Generative, Block-wise Matching=true2024.03 | 37.72 | 24.03 | 30.88 | |
| DiffusionITM +Paradigm=Zero-shot, Approach=Generative, Block-wise Matching=true2024.03 | 37.72 | 29.88 | 33.8 | |
| HN-DiffusionITM +Paradigm=Zero-shot, Approach=Generative, Block-wise Matching=true2024.03 | 37.55 | 30.37 | 33.96 | |
| Diffusion ClassifierParadigm=Zero-shot, Approach=Discriminative, Block-wise Matching=true2024.03 | 34.59 | 24.12 | 29.36 | |
| DiffusionITMParadigm=Zero-shot, Approach=Discriminative, Block-wise Matching=true2024.03 | 34.59 | 29.83 | 32.21 | |
| Random ChanceParadigm=Zero-shot, Approach=Discriminative2024.03 | 25 | 25 | 25 |