Image-to-Text Retrieval on Flickr8k
58.5R@1CLIP-PGS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CLIP-PGSImage Enc.=ViT-B/16, Ratio=0.5, Zero-shot=true2025.03 | 58.5 | 83.6 | 90.6 | |
| CLIP-PGSImage Enc.=ViT-B/16, Ratio=0.3, Zero-shot=true2025.03 | 58.3 | 82.9 | 90.8 | |
| E-CLIPImage Enc.=ViT-B/16, Zero-shot=true2025.03 | 57 | 82.7 | 90.1 | |
| CLIPImage Enc.=ViT-B/16, Zero-shot=true2025.03 | 55.7 | 81.6 | 89.9 | |
| FLIPImage Enc.=ViT-B/16, Zero-shot=true2025.03 | 55 | 80.9 | 88.9 | |
| A-CLIPImage Enc.=ViT-B/16, Zero-shot=true2025.03 | 53.7 | 80.1 | 88 | |
| CLIP-PGSImage Enc.=ViT-S/16, Ratio=0.5, Zero-shot=true2025.03 | 53.6 | 81.5 | 89.3 | |
| CLIP-PGSImage Enc.=ViT-S/16, Ratio=0.3, Zero-shot=true2025.03 | 52.3 | 80.9 | 89.7 | |
| Ours-Embeddingstrategy=Embedding, mode=Zero-shot2024.05 | 42 | 69.4 | 79.64 | |
| Ours-RGB0.3strategy=RGB, anchor_patch_ratio=0.3, mode=Zero-shot2024.05 | 40.9 | 68.22 | 79 | |
| Ours-RGB0.5strategy=RGB, anchor_patch_ratio=0.5, mode=Zero-shot2024.05 | 40.84 | 68.72 | 79.14 | |
| FLIPAttnmode=Zero-shot2024.05 | 40.58 | 68.88 | 78.86 | |
| CLIPmode=Zero-shot2024.05 | 40.54 | 68.9 | 80.2 | |
| FLIPmode=Zero-shot2024.05 | 40.32 | 68.1 | 78.64 | |
| Ours-Kmeansstrategy=Kmeans, mode=Zero-shot2024.05 | 40.1 | 69.86 | 79.91 | |
| CLIP-RefineBackbone=ViT-B/32, Zero-shot=true2025.04 | 38.28 | 61.55 | 70.94 | |
| HyCDBackbone=ViT-B/32, Zero-shot=true2025.04 | 34.69 | 57.19 | 67.22 | |
| HyCD + LalignBackbone=ViT-B/32, Zero-shot=true2025.04 | 32.58 | 56.94 | 66.99 | |
| Self-KDBackbone=ViT-B/32, Zero-shot=true2025.04 | 32.47 | 55.11 | 65.1 | |
| Pre-trained (CLIP)Backbone=ViT-B/32, Zero-shot=true2025.04 | 32.42 | 55.57 | 65.26 | |
| ContrastiveBackbone=ViT-B/32, Zero-shot=true2025.04 | 28.83 | 51.55 | 61.07 | |
| m²-mixBackbone=ViT-B/32, Zero-shot=true2025.04 | 28.64 | 50.65 | 60.77 |