Image-Text Retrieval on COCO
62.63Recall@1DetailCLIP-ViT-L/14
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DetailCLIP-ViT-L/14Single Feature=Yes, Input=CC@102022.08 | 62.63 | — | |
| CLIP-ViT-L/14Single Feature=No, Input=CC@102022.08 | 59.4 | — | |
| InternVL-C (MLLM)Single Feature=Yes, Input=Full Image2022.08 | 57.12 | — | |
| CLIP-ViT-L/14Single Feature=Yes, Input=Full Image2022.08 | 56.74 | — | |
| HQ-CLIPDataset (Data Size)=DataComp-DFN (130M), Samples Seen@Resolution=3.20B@224, GPU-hours=≈ 2675, Zero-shot=true2026.04 | 52.2 | — | |
| GLIP (detector/grounding-based)Single Feature=Yes, Input=Full Image2022.08 | 50.87 | — | |
| MDETR (detector/grounding-based)Single Feature=Yes, Input=Full Image2022.08 | 49.65 | — | |
| MetaCLIP-400MDataset (Data Size)=(400M), Samples Seen@Resolution=12.8B@224, GPU-hours=≈ 10700, Zero-shot=true2026.04 | 48.2 | — | |
| BLIP2 (CLIP-like)Single Feature=Yes, Input=Full Image2022.08 | 47.85 | — | |
| SLIP (CLIP-like)Single Feature=Yes, Input=Full Image2022.08 | 47.49 | — | |
| OpenCLIPDataset (Data Size)=LAION-400M (400M), Samples Seen@Resolution=12.8B@224, GPU-hours=10736, Zero-shot=true2026.04 | 46.9 | — | |
| DynamiCSDataset (Data Size)=LAION-400M (298M), Samples Seen@Resolution=2.56B@112 + 128M@224, GPU-hours=299, Zero-shot=true2026.04 | 46.5 | — | |
| DynamiCSDataset (Data Size)=DataComp-DFN (130M), Samples Seen@Resolution=2.56B@112 + 128M@224, GPU-hours=299, Zero-shot=true2026.04 | 46.2 | — | |
| RegionCLIP (detector-based)Single Feature=Yes, Input=Full Image2022.08 | 46.06 | — | |
| DynamiCSDataset (Data Size)=LAION-400M (298M), Samples Seen@Resolution=1.28B@112 + 128M@224, GPU-hours=163, Zero-shot=true2026.04 | 45.1 | — | |
| RECLIP*Dataset (Data Size)=LAION-400M (298M), Samples Seen@Resolution=2.56B@112 + 128M@224, GPU-hours=280, Zero-shot=true2026.04 | 44.4 | — | |
| CLIPA*Dataset (Data Size)=LAION-400M (298M), Samples Seen@Resolution=2.56B@112 + 128M@224, GPU-hours=269, Zero-shot=true2026.04 | 44.4 | — | |
| DFN*Dataset (Data Size)=DataComp-DFN (130M), Samples Seen@Resolution=1.28B@112 + 128M@224, GPU-hours=151, Zero-shot=true2026.04 | 44.3 | — | |
| DynamiCSDataset (Data Size)=DataComp-DFN (130M), Samples Seen@Resolution=1.28B@112 + 128M@224, GPU-hours=163, Zero-shot=true2026.04 | 44.2 | — | |
| DetailCLIP-ViT-B/16Single Feature=Yes, Input=CC@102022.08 | 44.19 | — | |
| OpenAI-WITDataset (Data Size)=-, Samples Seen@Resolution=12.8B@224, GPU-hours=10700, Zero-shot=true2026.04 | 42.8 | — | |
| RandomDataset (Data Size)=LAION-400M (298M), Samples Seen@Resolution=1.28B@112 + 128M@224, GPU-hours=151, Zero-shot=true2026.04 | 42.7 | — | |
| DynamiCSDataset (Data Size)=DataComp-DFN (130M), Samples Seen@Resolution=0.64B@112 + 128M@224, GPU-hours=95, Zero-shot=true2026.04 | 42.5 | — | |
| CLIP-ViT-B/16Single Feature=No, Input=CC@102022.08 | 41.24 | — | |
| CLIP-ViT-B/16Single Feature=Yes, Input=Full Image2022.08 | 40.93 | — | |
| DataCompDataset (Data Size)=DataComp-Large (1.28B), Samples Seen@Resolution=1.28B@224, GPU-hours=≈ 1070, Zero-shot=true2026.04 | 39.8 | — | |
| RandomDataset (Data Size)=DataComp-DFN (130M), Samples Seen@Resolution=0.64B@112 + 128M@224, GPU-hours=90, Zero-shot=true2026.04 | 39.6 | — | |
| ACED-F0Samples Seen=13B, Inference GFlops=3.3, Zero-shot=true2024.11 | — | 51 | |
| ACED-F1Samples Seen=13B, Inference GFlops=7.14, Zero-shot=true2024.11 | — | 55.6 | |
| ACED-F2Samples Seen=13B, Inference GFlops=10.29, Zero-shot=true2024.11 | — | 58.3 | |
| CLIP-KD-RN50Samples Seen=0.5B, Inference GFlops=9.09, Zero-shot=true2024.11 | — | 43.5 | |
| DataComp-CLIP-B/32Samples Seen=13B, Inference GFlops=7.39, Zero-shot=true2024.11 | — | 45.4 | |
| DatologyAI-cls-B/32Samples Seen=5.1B, Inference GFlops=7.39, Zero-shot=true2024.11 | — | 38.5 | |
| DatologyAI-cls-S/32Samples Seen=2.0B, Inference GFlops=2.83, Zero-shot=true2024.11 | — | 30.2 | |
| DatologyAI-ret-B/32Samples Seen=5.1B, Inference GFlops=7.39, Zero-shot=true2024.11 | — | 49.6 | |
| DatologyAI-ret-S/32Samples Seen=2.0B, Inference GFlops=2.83, Zero-shot=true2024.11 | — | 41.5 | |
| DreamLIP2025.09 | — | 52.7 | |
| LAION-CLIP-B/32Samples Seen=34B, Inference GFlops=7.39, Zero-shot=true2024.11 | — | 47.7 | |
| LSSBackbone=ViT-B/162025.09 | — | 51.8 | |
| MetaCLIP-CLIP-B/32Samples Seen=13B, Inference GFlops=7.39, Zero-shot=true2024.11 | — | 46.7 | |
| MobileCLIP-S0Samples Seen=13B, Inference GFlops=3.7, Zero-shot=true2024.11 | — | 49.6 | |
| MobileCLIP-S1Samples Seen=13B, Inference GFlops=7.64, Zero-shot=true2024.11 | — | 53 | |
| MobileCLIP-S2Samples Seen=13B, Inference GFlops=10.81, Zero-shot=true2024.11 | — | 54.4 | |
| OpenAI-CLIP-B/32Samples Seen=13B, Inference GFlops=7.39, Zero-shot=true2024.11 | — | 40.3 | |
| OpenAI-RN101Samples Seen=13B, Inference GFlops=12.75, Zero-shot=true2024.11 | — | 40.3 | |
| OpenAI-RN50Samples Seen=13B, Inference GFlops=9.09, Zero-shot=true2024.11 | — | 38.7 | |
| SigLIP2025.09 | — | 56.8 | |
| TinyCLIP-39M/16Samples Seen=20B, Inference GFlops=9.48, Zero-shot=true2024.11 | — | 46.9 | |
| TinyCLIP-45M/32Samples Seen=15.8B, Inference GFlops=3.7, Zero-shot=true2024.11 | — | 45.4 | |
| TinyCLIP-63M/32Samples Seen=15.8B, Inference GFlops=5.65, Zero-shot=true2024.11 | — | 47.7 | |
| TinyCLIP-RN30MSamples Seen=15.2B, Inference GFlops=6.93, Zero-shot=true2024.11 | — | 43.3 |