Compositional Image Retrieval on FashionIQ 1.0 (val)
43.49Average Recall@10BLIP4CIR
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| BLIP4CIRBackbone=BLIP, Supervision=Fully-supervised2025.01 | 43.49 | 67.31 | 42.09 | 67.33 | 41.76 | 64.28 | 46.61 | 70.32 | |
| SCOTBackbone=BLIP-2, Supervision=Zero-Shot2025.01 | 38.45 | 60.03 | 32.78 | 55.91 | 41.42 | 61.09 | 41.15 | 63.1 | |
| CLIP4CIRBackbone=CLIP L/14, Supervision=Fully-supervised2025.01 | 38.32 | 61.74 | 33.81 | 59.4 | 39.99 | 60.45 | 41.41 | 65.37 | |
| CoLLMVision Encoder=BLIP-L/16, Evaluation Protocol=Zero-shot, Incorporates LLM=true2025.03 | 34.6 | 56 | — | — | — | — | — | — | |
| FashionVLPBackbone=Multi, Supervision=Fully-supervised2025.01 | 34.27 | 62.51 | 32.42 | 60.29 | 31.89 | 58.44 | 38.51 | 68.79 | |
| Slerp-TATVision Encoder=BLIP-L/16, Evaluation Protocol=Zero-shot2025.03 | 32.8 | 53.3 | — | — | — | — | — | — | |
| SCOTBackbone=BLIP, Supervision=Zero-Shot2025.01 | 30.68 | 51.33 | 26.42 | 49.23 | 30.91 | 49.65 | 34.72 | 55.12 | |
| DCNetBackbone=Multi, Supervision=Fully-supervised2025.01 | 30.44 | 58.29 | 28.95 | 56.07 | 23.95 | 47.3 | 30.44 | 58.29 | |
| CoLLMVision Encoder=OpenAI CLIP-L/14, Evaluation Protocol=Zero-shot, Incorporates LLM=true2025.03 | 30.1 | 49.5 | — | — | — | — | — | — | |
| Image+TextBackbone=BLIP-2, Supervision=Zero-Shot2025.01 | 29.21 | 50.05 | 23.3 | 45.61 | 32.82 | 53.09 | 31.51 | 51.45 | |
| CIREVLVision Encoder=OpenAI CLIP-L/14, Evaluation Protocol=Zero-shot, Incorporates LLM=true2025.03 | 28.6 | 48.6 | — | — | — | — | — | — | |
| TransAggBackbone=CLIP L/14, Supervision=Zero-Shot2025.01 | 28.57 | 48.29 | 23.85 | 44.57 | 29.54 | 47.79 | 32.33 | 52.52 | |
| CIREVLVision Encoder=OpenAI CLIP-B/32, Evaluation Protocol=Zero-shot, Incorporates LLM=true2025.03 | 28.3 | 49.4 | — | — | — | — | — | — | |
| Slerp-TATVision Encoder=OpenAI CLIP-L/14, Evaluation Protocol=Zero-shot2025.03 | 28.3 | 47.6 | — | — | — | — | — | — | |
| SCOTBackbone=CLIP L/14, Supervision=Zero-Shot2025.01 | 28.27 | 47.44 | 23.69 | 45.06 | 29.09 | 47.01 | 32.02 | 50.33 | |
| ContextI2WVision Encoder=OpenAI CLIP-L/14, Evaluation Protocol=Zero-shot2025.03 | 27.8 | 48.9 | — | — | — | — | — | — | |
| TransAggBackbone=BLIP, Supervision=Zero-Shot2025.01 | 26.95 | 46.1 | 21.67 | 41.89 | 28.07 | 45.63 | 31.11 | 50.79 | |
| LinCIRVision Encoder=OpenAI CLIP-L/14, Evaluation Protocol=Zero-shot, Reproduced Results=true2025.03 | 26.4 | 46.6 | — | — | — | — | — | — | |
| SEARLEVision Encoder=OpenAI CLIP-L/14, Evaluation Protocol=Zero-shot2025.03 | 25.6 | 46.2 | — | — | — | — | — | — | |
| SEARLE-XLBackbone=CLIP L/14, Supervision=Zero-Shot2025.01 | 25.56 | 46.23 | 20.48 | 43.13 | 26.89 | 45.58 | 29.32 | 49.97 | |
| CoLLMVision Encoder=OpenAI CLIP-B/32, Evaluation Protocol=Zero-shot, Incorporates LLM=true2025.03 | 24.8 | 45.2 | — | — | — | — | — | — | |
| Pic2WordBackbone=CLIP L/14, Supervision=Zero-Shot2025.01 | 24.7 | 43.7 | 20 | 40.2 | 26.2 | 43.6 | 27.9 | 47.4 | |
| Pic2WorldVision Encoder=OpenAI CLIP-L/14, Evaluation Protocol=Zero-shot2025.03 | 24.7 | 43.7 | — | — | — | — | — | — | |
| Text-OnlyBackbone=BLIP-2, Supervision=Zero-Shot2025.01 | 24.68 | 43.59 | 20.77 | 41.64 | 25.95 | 42.83 | 27.33 | 46.31 | |
| MAAFBackbone=Multi, Supervision=Fully-supervised2025.01 | 24.3 | 48.8 | 23.8 | 48.6 | 21.3 | 44.2 | 27.9 | 53.6 | |
| SCOTBackbone=CLIP B/32, Supervision=Zero-Shot2025.01 | 24.14 | 43.44 | 19.73 | 41.24 | 25.51 | 42.93 | 27.18 | 46.14 | |
| Text-OnlyBackbone=BLIP, Supervision=Zero-Shot2025.01 | 24.01 | 42.73 | 20.03 | 39.96 | 24.63 | 41.02 | 27.38 | 47.22 | |
| TransAggBackbone=CLIP B/32, Supervision=Zero-Shot2025.01 | 23.91 | 44.68 | 19.44 | 42.04 | 25.37 | 42.69 | 26.93 | 49.31 | |
| Slerp-TATVision Encoder=OpenAI CLIP-B/32, Evaluation Protocol=Zero-shot2025.03 | 23 | 44 | — | — | — | — | — | — | |
| SEARLEVision Encoder=OpenAI CLIP-B/32, Evaluation Protocol=Zero-shot2025.03 | 22.9 | 42.5 | — | — | — | — | — | — | |
| SEARLEBackbone=CLIP B/32, Supervision=Zero-Shot2025.01 | 22.89 | 42.53 | 18.54 | 39.51 | 24.44 | 41.61 | 18.54 | 39.51 | |
| PALAVRAVision Encoder=OpenAI CLIP-B/32, Evaluation Protocol=Zero-shot2025.03 | 19.8 | 37.3 | — | — | — | — | — | — | |
| PALAVRABackbone=CLIP B/32, Supervision=Zero-Shot2025.01 | 19.76 | 37.25 | 17.25 | 35.94 | 21.49 | 37.05 | 20.55 | 38.76 | |
| Text-OnlyBackbone=CLIP L/14, Supervision=Zero-Shot2025.01 | 19.01 | 35.26 | 15.22 | 33.01 | 19.82 | 33.31 | 21.87 | 39.46 | |
| Text-OnlyBackbone=CLIP B/32, Supervision=Zero-Shot2025.01 | 18.41 | 36.28 | 14.92 | 33.81 | 19.77 | 34.69 | 20.55 | 40.33 | |
| Image+TextBackbone=CLIP L/14, Supervision=Zero-Shot2025.01 | 18.12 | 33.17 | 14.27 | 31.33 | 19.13 | 32.28 | 20.95 | 35.9 | |
| Image+TextBackbone=CLIP B/32, Supervision=Zero-Shot2025.01 | 13.36 | 27.51 | 12.44 | 28.55 | 12.61 | 24.82 | 15.04 | 29.16 | |
| Image+TextBackbone=BLIP, Supervision=Zero-Shot2025.01 | 8.06 | 18.16 | 6.14 | 19.78 | 9.37 | 19.87 | 8.66 | 19.78 | |
| Image-OnlyBackbone=CLIP L/14, Supervision=Zero-Shot2025.01 | 7.97 | 17.43 | 5.25 | 13.63 | 10.54 | 20.65 | 8.1 | 18.01 | |
| Image-OnlyBackbone=BLIP-2, Supervision=Zero-Shot2025.01 | 7.53 | 17.93 | 4.21 | 11.89 | 10.59 | 23.51 | 7.81 | 18.41 | |
| Image-OnlyBackbone=BLIP, Supervision=Zero-Shot2025.01 | 6.65 | 15.4 | 5.05 | 12.19 | 7.55 | 17.76 | 7.34 | 16.26 | |
| Image-OnlyBackbone=CLIP B/32, Supervision=Zero-Shot2025.01 | 5.88 | 13.19 | 6.96 | 14.08 | 4.46 | 11.89 | 6.22 | 13.61 |