Image Retrieval on Cars (R@1, R@2, R@4)
98.2R@1UNICOM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| UNICOMBackbone=ViT-L/14, Fine-tuning=true, Resolution=3362023.04 | 98.2 | — | — | |
| UNICOMBackbone=ViT-L/14, Fine-tuning=true, Resolution=2242023.04 | 97.9 | — | — | |
| UNICOMBackbone=ViT-B/16, Fine-tuning=true, Resolution=2242023.04 | 97.7 | — | — | |
| UNICOMBackbone=ViT-L/14, Resolution=336x3362023.04 | 97.3 | — | — | |
| UNICOMBackbone=ViT-B/32, Fine-tuning=true, Resolution=2242023.04 | 97.3 | — | — | |
| UNICOMBackbone=ViT-L/142023.04 | 96.9 | — | — | |
| UNICOMBackbone=ViT-B/162023.04 | 96.8 | — | — | |
| UNICOMBackbone=ViT-B/322023.04 | 95.9 | — | — | |
| OpenCLIPBackbone=ViT-H/142023.04 | 95.7 | — | — | |
| OpenCLIPBackbone=ViT-L/142023.04 | 94.9 | — | — | |
| SE-ResNet-50Backbone=SE-ResNet-502023.04 | 94.8 | — | — | |
| OpenCLIPBackbone=ViT-B/162023.04 | 92.9 | — | — | |
| HIERbackbone=DINO, embedding dimension=3842022.12 | 91.3 | 95.2 | 97.1 | |
| CLIPBackbone=ViT-L/14, Resolution=336x3362023.04 | 90.9 | — | — | |
| CLIPBackbone=ViT-L/142023.04 | 90.3 | — | — | |
| Hypbackbone=DINO, embedding dimension=3842022.12 | 89.2 | 94.1 | 96.7 | |
| OpenCLIPBackbone=ViT-B/322023.04 | 89.2 | — | — | |
| HIERbackbone=DeiT, embedding dimension=3842022.12 | 88.9 | 93.9 | 96.6 | |
| HIERbackbone=DINO, embedding dimension=1282022.12 | 88.4 | 93.3 | 95.9 | |
| HIERbackbone=ViT, embedding dimension=3842022.12 | 88.3 | 93.2 | 96.1 | |
| HIERbackbone=ResNet-50, embedding dimension=5122022.12 | 88.2 | 93 | 95.6 | |
| ProxyNCA++backbone=ResNet-50, embedding dimension=512, larger input images=true2022.12 | 86.5 | 92.5 | 95.7 | |
| Hypbackbone=ViT, embedding dimension=3842022.12 | 86.5 | 92.1 | 95.3 | |
| HIERbackbone=ViT, embedding dimension=1282022.12 | 86.4 | 91.9 | 95.1 | |
| Hypbackbone=DeiT, embedding dimension=3842022.12 | 86.4 | 92.2 | 95.5 | |
| PAbackbone=Inception with BatchNorm, embedding dimension=5122022.12 | 86.1 | 91.7 | 95 | |
| Hypbackbone=DINO, embedding dimension=1282022.12 | 86 | 91.9 | 95.2 | |
| CLIPBackbone=ViT-B/162023.04 | 85.2 | — | — | |
| HIERbackbone=DeiT, embedding dimension=1282022.12 | 85.1 | 91.1 | 95.1 | |
| SoftTriplebackbone=Inception with BatchNorm, embedding dimension=5122022.12 | 84.5 | 90.7 | 94.5 | |
| NSoftmaxbackbone=ResNet-50, embedding dimension=5122022.12 | 84.2 | 90.4 | 94.4 | |
| MSbackbone=Inception with BatchNorm, embedding dimension=5122022.12 | 84.1 | 90.4 | 94 | |
| Hypbackbone=ViT, embedding dimension=1282022.12 | 82.7 | 89.7 | 93.9 | |
| MICbackbone=ResNet-50, embedding dimension=1282022.12 | 82.6 | 89.1 | 93.2 | |
| Hypbackbone=DeiT, embedding dimension=1282022.12 | 82.1 | 89.1 | 93.4 | |
| XBMbackbone=Inception with BatchNorm, embedding dimension=5122022.12 | 82 | 88.7 | 93.1 | |
| Hypbackbone=ResNet-50, embedding dimension=5122022.12 | 81.9 | 88.8 | 93.1 | |
| NSoftmaxbackbone=ResNet-50, embedding dimension=1282022.12 | 81.6 | 88.7 | 93.4 | |
| HTLbackbone=Inception with BatchNorm, embedding dimension=5122022.12 | 81.4 | 88 | 92.7 | |
| XBMbackbone=ResNet-50, embedding dimension=1282022.12 | 80.6 | 91.6 | 96.2 | |
| CLIPBackbone=ViT-B/322023.04 | 79 | — | — | |
| DeiT-Sbackbone=DeiT, embedding dimension=3842022.12 | 52.8 | 65.1 | 76.2 | |
| ViT-Sbackbone=ViT, embedding dimension=3842022.12 | 47.8 | 60.2 | 72.2 | |
| DINObackbone=DINO, embedding dimension=3842022.12 | 42.9 | 53.9 | 64.2 |