List your company

Image Retrieval on CUB

90.1Recall@1UNICOM
40.753.52566.3579.175Apr 8, 2019Dec 8, 2019Aug 8, 2020Apr 9, 2021Dec 9, 2021Aug 10, 2022Apr 12, 2023

Evaluation results

MethodLinks
UNICOMBackbone=ViT-L/14, Fine-tuning=true, Resolution=3362023.0490.1————
UNICOMBackbone=ViT-L/14, Fine-tuning=true, Resolution=2242023.0489.7————
UNICOMBackbone=ViT-L/14, Resolution=336x3362023.0489.2————
UNICOMBackbone=ViT-B/16, Fine-tuning=true, Resolution=2242023.0488.8————
UNICOMBackbone=ViT-L/142023.0488.5————
UNICOMBackbone=ViT-B/162023.0486.5————
UNICOMBackbone=ViT-B/32, Fine-tuning=true, Resolution=2242023.0485.8————
HIERbackbone=ViT, embedding dimension=3842022.1285.791.394.4——
Hypbackbone=ViT, embedding dimension=3842022.1285.691.494.8——
Hyp-ViTBackbone=ViT-S/162023.0485.6————
HIERbackbone=ViT, embedding dimension=1282022.1284.290.193.7——
Hypbackbone=ViT, embedding dimension=1282022.128490.294.2——
UNICOMBackbone=ViT-B/322023.0483.7————
ViT-Sbackbone=ViT, embedding dimension=3842022.1283.190.494.4——
OpenCLIPBackbone=ViT-H/142023.0483.1————
HIERbackbone=DINO, embedding dimension=3842022.1281.188.293.3——
Hypbackbone=DINO, embedding dimension=3842022.1280.987.692.4——
OpenCLIPBackbone=ViT-L/142023.0479.4————
CRTEmbedding Dimension=1282022.1078.9886.6891.6195.04—
HIERbackbone=DeiT, embedding dimension=3842022.1278.786.892——
HIERbackbone=DINO, embedding dimension=1282022.1278.586.791.5——
Hypbackbone=DINO, embedding dimension=1282022.1278.38691.2——
Hypbackbone=DeiT, embedding dimension=3842022.1277.886.691.9——
CLIPBackbone=ViT-L/14, Resolution=336x3362023.0477.3————
IRTRbackbone=DeiT, embedding dimension=3842022.1276.68591.1——
CLIPBackbone=ViT-L/142023.0476————
HIERbackbone=DeiT, embedding dimension=1282022.1275.284.290——
IRTREmbedding Dimension=3842022.1074.782.989.393.3—
Hyp-DeiTEmbedding Dimension=1282022.1074.784.590.194.1—
Hypbackbone=DeiT, embedding dimension=1282022.1274.784.590.1——
Group Loss++Embedding Dimension=5122022.1072.680.586.291.2—
IRTRbackbone=DeiT, embedding dimension=1282022.1272.681.988.7——
ETLREmbedding Dimension=5122022.1072.181.387.6——
OpenCLIPBackbone=ViT-B/162023.0471.4————
DINObackbone=DINO, embedding dimension=3842022.1270.881.188.8——
DeiT-Sbackbone=DeiT, embedding dimension=3842022.1270.681.388.7——
IBCEmbedding Dimension=5122022.1070.380.387.692.7—
HIERbackbone=ResNet-50, embedding dimension=5122022.1270.179.486.9——
ProxyNCA++backbone=ResNet-50, embedding dimension=512, larger input images=true2022.126979.887.3——
DRML-PAEmbedding Dimension=5122022.1068.778.686.391.6—
LSCM-GNNEmbedding Dimension=5122022.1068.577.385.391.3—
Proxy-AnchorEmbedding Dimension=5122022.1068.479.286.891.6—
DCML-MDWEmbedding Dimension=5122022.1068.477.986.191.7—
D & CEmbedding Dimension=5122022.1068.478.78691.6—
PAbackbone=Inception with BatchNorm, embedding dimension=5122022.1268.479.286.8——
PA+DIMLEmbedding Dimension=1282022.1066.46————
MICbackbone=ResNet-50, embedding dimension=1282022.1266.176.885.6——
CLIPBackbone=ViT-B/162023.0466.1————
XBMbackbone=Inception with BatchNorm, embedding dimension=5122022.1265.875.984——
MSEmbedding Dimension=5122022.1065.77786.391.2—
MSbackbone=Inception with BatchNorm, embedding dimension=5122022.1265.77786.3——
Hypbackbone=ResNet-50, embedding dimension=5122022.1265.576.284.9——
SoftTriplebackbone=Inception with BatchNorm, embedding dimension=5122022.1265.476.484.5——
EPSHNEmbedding Dimension=5122019.0464.975.383.5——
DREMLEmbedding Dimension=5762019.0463.97583.1——
Direct Gradient FrameworkEmbedding Dimension=64, Batch Size=2562022.0163.874.883.7——
Direct Gradient FrameworkEmbedding Dimension=64, Batch Size=3842022.0163.87584.2——
Direct Gradient FrameworkEmbedding Dimension=64, Batch Size=1282022.0163.574.883.6——
Direct Gradient FrameworkEmbedding Dimension=64, Batch Size=5122022.0163.174.683.2——
OpenCLIPBackbone=ViT-B/322023.0462.3————
Proxy-anchorEmbedding Dimension=642022.0161.77381.8——
NSoftmaxbackbone=ResNet-50, embedding dimension=5122022.1261.373.983.5——
DR-MSEmbedding Dimension=64, Batch Size=128, Implementation=Paper Re-implementation2022.0160.771.981.3——
ABEEmbedding Dimension=5122019.0460.671.579.8——
SoftTripleEmbedding Dimension=642022.0160.171.981.2——
MSEmbedding Dimension=64, Batch Size=128, Implementation=Paper Re-implementation2022.0159.871.781——
DR-MSEmbedding Dimension=642022.0159.17180.3——
SCTBackbone=ResNet-50, Embedding Dimension=642020.0757.769.879.687—
SCTEmbedding Dimension=642022.0157.769.879.6——
A-BIEREmbedding Dimension=5122022.1057.568.778.382.6—
MSEmbedding Dimension=642022.0157.469.880——
EasyPositiveEmbedding Dimension=642022.0157.368.979.3——
HTLEmbedding Dimension=5122019.0457.168.878.7——
HTLbackbone=Inception with BatchNorm, embedding dimension=5122022.1257.168.878.7——
SHNBackbone=ResNet-50, Embedding Dimension=642020.0756.768.678.686.6—
CLIPBackbone=ViT-B/322023.0456.7————
NSoftmaxbackbone=ResNet-50, embedding dimension=1282022.1256.569.679.9——
BIEREmbedding Dimension=5122019.0455.367.276.9——
HDCEmbedding Dimension=3842019.0453.665.777——
N-pairBackbone=ResNet-50, Embedding Dimension=642020.0751.964.374.983.2—
ClusteringBackbone=ResNet-50, Embedding Dimension=642020.0749.861.471.881.9—
SmartMiningBackbone=ResNet-50, Embedding Dimension=642020.0749.862.374.183.3—
ProxyNCABackbone=ResNet-50, Embedding Dimension=642020.0749.261.967.972.4—
ProxyNCAEmbedding Dimension=642022.0149.261.967.9——
LiftedBackbone=ResNet-50, Embedding Dimension=642020.0743.656.668.679.6—
LiftedStructEmbedding Dimension=642022.0143.656.668.6——
Triplet-SemihardBackbone=ResNet-50, Embedding Dimension=642020.0742.65566.477.2—
CircleNoisy Label Rate=25%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————44.07
CircleNoisy Label Rate=50%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————22.96
Co-teachingNoisy Label Rate=25%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————51.75
Co-teachingNoisy Label Rate=50%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————48.85
Co-teaching w/ TemperatureNoisy Label Rate=25%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————54.59
Co-teaching w/ TemperatureNoisy Label Rate=50%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————48.32
Co-teaching+Noisy Label Rate=25%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————51.55
Co-teaching+Noisy Label Rate=50%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————47.62
Contrastive LossNoisy Label Rate=25%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————47.27
Contrastive LossNoisy Label Rate=50%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————39.43
FastAPNoisy Label Rate=25%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————31.18
FastAPNoisy Label Rate=50%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————40.83
MCL + PRISMNoisy Label Rate=25%, Label Noise Type=Small Cluster, Backbone=BN-Inception, Output Dimension=5122021.03————55.77