Fine-grained Image Classification on Stanford Cars
95.7AccuracyCAP
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CAPTraining data setup=primary2021.01 | 95.7 | — | — | — | — | — | |
| CCFRBase Model=ResNet-50, re-ranking=true2021.02 | 95.49 | — | — | — | — | — | |
| CCFRBase Model=ResNet-50, re-ranking=false2021.02 | 95.37 | — | — | — | — | — | |
| API-NetBackbone=DenseNet-161, Extra Supervision=No2020.02 | 95.3 | — | — | — | — | — | |
| API-NetBackbone=DenseNet-1612021.03 | 95.3 | — | — | — | — | — | |
| PMGBase Model=ResNet50, Evaluation Mode=Combined Accuracy2020.03 | 95.1 | — | — | — | — | — | |
| PMGBase Model=ResNet-502021.02 | 95.1 | — | — | — | — | — | |
| PMGBackbone=ResNet-502021.03 | 95.1 | — | — | — | — | — | |
| PMGBase Model=ResNet502020.03 | 95 | — | — | — | — | — | |
| Ours (multi scale)Backbone=ResNet-50, Resolution=448, #Parameters=23.9M, scale_mode=multi2019.12 | 94.9 | — | — | — | — | — | |
| API-NetBackbone=ResNet-101, Extra Supervision=No2020.02 | 94.9 | — | — | — | — | — | |
| Mix+Base Model=ResNet-502021.02 | 94.9 | — | — | — | — | — | |
| API-NetBackbone=ResNet-101, Input size=448 x 4482021.01 | 94.9 | — | — | — | — | — | |
| API-NetBackbone=ResNet-50, Extra Supervision=No2020.02 | 94.8 | — | — | — | — | — | |
| AutoAugTraining data setup=primary + secondary2021.01 | 94.8 | — | — | — | — | — | |
| DF-GMMBase Model=ResNet-502021.02 | 94.8 | — | — | — | — | — | |
| API-netBase Model=ResNet-502021.02 | 94.8 | — | — | — | — | — | |
| TransFGBackbone=ViT-B_162021.03 | 94.8 | — | — | — | — | — | |
| API-NetBackbone=ResNet-502020.06 | 94.8 | — | — | — | — | 2.3 | |
| API-NetBackbone=ResNet-50, Input size=448 x 4482021.01 | 94.8 | — | — | — | — | — | |
| S3NBase Model=ResNet502020.03 | 94.7 | — | — | — | — | — | |
| S3NBackbone=ResNet-50, Multi-cropping operations=true2020.06 | 94.7 | — | — | — | — | 1.7 | |
| GPipeTraining data setup=primary2021.01 | 94.6 | — | — | — | — | — | |
| ACNetBase Model=ResNet-502021.02 | 94.6 | — | — | — | — | — | |
| Cross-XBackbone=ResNet-502021.03 | 94.6 | — | — | — | — | — | |
| Cross-XBackbone=ResNet-502020.06 | 94.6 | — | — | — | — | 2.8 | |
| Cross-XBackbone=ResNet-50, Input size=448 x 4482021.01 | 94.6 | — | — | — | — | — | |
| ACNetBackbone=ResNet-50, Input size=448 x 4482021.01 | 94.6 | — | — | — | — | — | |
| PCA-NetBackbone=ResNet-101, Input size=448 x 4482021.01 | 94.6 | — | — | — | — | — | |
| DCLBase Model=ResNet502020.03 | 94.5 | — | — | — | — | — | |
| WS-DAN2020.10 | 94.5 | — | — | — | — | — | |
| DATL + WS-DAN2020.10 | 94.5 | — | — | — | — | — | |
| DCLTraining data setup=primary2021.01 | 94.5 | — | — | — | — | — | |
| DCLBase Model=ResNet-502021.02 | 94.5 | — | — | — | — | — | |
| DBTNetBackbone=ResNet-1012021.03 | 94.5 | — | — | — | — | — | |
| DCLBackbone=ResNet-502020.06 | 94.5 | — | — | — | — | 2.7 | |
| CINBackbone=ResNet-101, Input size=448 x 4482021.01 | 94.5 | — | — | — | — | — | |
| MC LossTraining data setup=primary2021.01 | 94.4 | — | — | — | — | — | |
| MC-LossBase Model=B-CNN, Model Component=2B+A2020.02 | 94.4 | — | — | — | — | — | |
| Ours (single scale)Backbone=ResNet-50, Resolution=448, #Parameters=23.9M, scale_mode=single2019.12 | 94.3 | — | — | — | — | — | |
| PMGBase Model=VGG16, Evaluation Mode=Combined Accuracy2020.03 | 94.3 | — | — | — | — | — | |
| BARMTraining data setup=primary2021.01 | 94.3 | — | — | — | — | — | |
| BCNBase Model=ResNet-502021.02 | 94.3 | — | — | — | — | — | |
| PCA-NetBackbone=ResNet-50, Input size=448 x 4482021.01 | 94.3 | — | — | — | — | — | |
| PMGBase Model=VGG162020.03 | 94.2 | — | — | — | — | — | |
| FDLBackbone=DenseNet-1612021.03 | 94.2 | — | — | — | — | — | |
| DLA2020.10 | 94.1 | — | — | — | — | — | |
| DBT-NetBackbone=ResNet-502020.06 | 94.1 | — | — | — | — | 5.7 | |
| CINBackbone=ResNet-50, Input size=448 x 4482021.01 | 94.1 | — | — | — | — | — | |
| GCLBase Model=ResNet-502021.02 | 94 | — | — | — | — | — | |
| SEFBackbone=ResNet-502020.06 | 94 | — | — | — | — | 4.8 | |
| PMGBackbone=DenseNet1612026.03 | 94 | — | — | — | — | — | |
| MaxEntBackbone=ResNet-50, #Parameters=23.9M2019.12 | 93.9 | — | — | — | — | — | |
| NTS-netBackbone=ResNet-50, Resolution=448, #Parameters=25.5M2019.12 | 93.9 | — | — | — | — | — | |
| NTS-NetBackbone=ResNet-50, Extra Supervision=No2020.02 | 93.9 | — | — | — | — | — | |
| MaxEntBackbone=DenseNet-161, Extra Supervision=No2020.02 | 93.9 | — | — | — | — | — | |
| NTS-NetBase Model=ResNet502020.03 | 93.9 | — | — | — | — | — | |
| MGE-CNNBase Model=ResNet502020.03 | 93.9 | — | — | — | — | — | |
| NTSBase Model=ResNet-502021.02 | 93.9 | — | — | — | — | — | |
| MGEBase Model=ResNet-502021.02 | 93.9 | — | — | — | — | — | |
| NTS-NetBase Model=ResNet50, Model Component=D+3A+6Conv.(3,3)2020.02 | 93.9 | — | — | — | — | — | |
| NTS-NetBackbone=ResNet-502021.03 | 93.9 | — | — | — | — | — | |
| DeiTBackbone=DeiT-B2021.03 | 93.9 | — | — | — | — | — | |
| NTS-NetBackbone=ResNet-50, Multi-cropping operations=true2020.06 | 93.9 | — | — | — | — | 6 | |
| NTS-NetBackbone=ResNet-50, Input size=448 x 4482021.01 | 93.9 | — | — | — | — | — | |
| TASNBackbone=ResNet-50, Resolution=448, #Parameters=35.2M2019.12 | 93.8 | — | — | — | — | — | |
| DFL-CNNExtra Supervision=No2020.02 | 93.8 | — | — | — | — | — | |
| DFL-CNN2020.10 | 93.8 | — | — | — | — | — | |
| DFL-CNNBase Model=VGG16, Model Component=B+2A+Conv.(1,1)2020.02 | 93.8 | — | — | — | — | — | |
| TASNBase Model=ResNet50, Model Component=D+A2020.02 | 93.8 | — | — | — | — | — | |
| DFB-CNNBackbone=ResNet-50, Separated initialization=true2020.06 | 93.8 | — | — | — | — | 6.3 | |
| TASNBackbone=ResNet-50, Multi-cropping operations=true2020.06 | 93.8 | — | — | — | — | 5 | |
| TASNBackbone=ResNet-50, Input size=448 x 4482021.01 | 93.8 | — | — | — | — | — | |
| HBPExtra Supervision=No2020.02 | 93.7 | — | — | — | — | — | |
| MC-LossBase Model=ResNet502020.03 | 93.7 | — | — | — | — | — | |
| MC-LossBase Model=ResNet50, Model Component=D+A2020.02 | 93.7 | — | — | — | — | — | |
| ViTBackbone=ViT-B_162021.03 | 93.7 | — | — | — | — | — | |
| MC-LossBackbone=ResNet-50, Input size=448 x 4482021.01 | 93.7 | — | — | — | — | — | |
| PMGBackbone=ResNet502026.03 | 93.6 | — | — | — | — | — | |
| Subset B2020.10 | 93.5 | — | — | — | — | — | |
| Inception-v3Base Model=Inception-v32021.02 | 93.5 | — | — | — | — | — | |
| PCBackbone=ResNet-50, #Parameters=23.9M2019.12 | 93.4 | — | — | — | — | — | |
| RA-CNNBackbone=DenseNet1612026.03 | 93.4 | — | — | — | — | — | |
| iSQRT-COVBackbone=ResNet-101, Extra Supervision=No2020.02 | 93.3 | — | — | — | — | — | |
| MPN-COV2020.10 | 93.3 | — | — | — | — | — | |
| Deep KSPDExtra Supervision=No2020.02 | 93.2 | — | — | — | — | — | |
| RAMBackbone=ResNet-50, Resolution=448, #Parameters=>23.9M2019.12 | 93.1 | — | — | — | — | — | |
| DFL-CNNBackbone=ResNet-50, Resolution=448, #Parameters=26.3M2019.12 | 93.1 | — | — | — | — | — | |
| DFLBase Model=ResNet502020.03 | 93.1 | — | — | — | — | — | |
| DFL-CNNBase Model=ResNet50, Model Component=D+2A+Conv.(1,1)2020.02 | 93.1 | — | — | — | — | — | |
| RAMBackbone=ResNet-50, Input size=448 x 4482021.01 | 93.1 | — | — | — | — | — | |
| DFL-CNNBackbone=ResNet-50, Input size=448 x 4482021.01 | 93.1 | — | — | — | — | — | |
| MAMCBackbone=ResNet-50, Resolution=448, #Parameters=434M2019.12 | 93 | — | — | — | — | — | |
| MAMCExtra Supervision=No2020.02 | 93 | — | — | — | — | — | |
| MAMC-CNNBackbone=ResNet-502020.06 | 93 | — | — | — | — | 7.7 | |
| MAMCBackbone=ResNet-101, Input size=448 x 4482021.01 | 93 | — | — | — | — | — | |
| PCExtra Supervision=No2020.02 | 92.9 | — | — | — | — | — | |
| PCBase Model=DenseNet1612020.03 | 92.9 | — | — | — | — | — | |
| PCBase Model=DenseNet161, Model Component=2E + 2A2020.02 | 92.9 | — | — | — | — | — | |
| ResNet-50Backbone=ResNet-50, Multi-cropping operations=true2020.06 | 92.9 | — | — | — | — | 8.3 |