Image Classification on FGVC-Aircraft (test)
95.4AccuracySR-GNN (Full Model)
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SR-GNN (Full Model)Param (M)=30.9, GFLOPS (B)=9.8, Infer. time per img. (ms)=5.02022.09 | 95.4 | — | — | — | — | 0.5 | — | — | |
| Ensemblemode=ensemble2021.07 | 95.11 | — | — | — | — | — | — | — | |
| CAPParam (M)=34.2, GFLOPS (B)=10.2, Infer. time per img. (ms)=4.22022.09 | 94.9 | — | — | — | — | — | — | — | |
| PARTType=Joint, Backbone=ResNet1012022.12 | 94.6 | — | — | — | — | — | — | — | |
| Non-binary Transfer Learning (Optimal)reinitialized layers=3 new layers, FC learning rate=0.01, high learning rate=0.025, low learning rate=0.01, low layers count=8, regularization=none2021.07 | 94.5 | — | — | — | — | — | — | — | |
| PARTType=Joint, Backbone=ResNet502022.12 | 94.4 | — | — | — | — | — | — | — | |
| State of the artresolution=448x4482021.07 | 93.9 | — | — | — | — | — | — | — | |
| L2 (Default)learning rate=0.025, regularization=L22021.07 | 93.78 | — | — | — | — | — | — | — | |
| SCAPNetInfer. time per img. (ms)=24.42022.09 | 93.6 | — | — | — | — | — | — | — | |
| OursBase Model=ResNet502021.01 | 93.42 | — | — | — | — | — | — | — | |
| PMGType=Feature Based, Backbone=ResNet502022.12 | 93.4 | — | — | — | — | — | — | — | |
| API-NetType=Feature Based, Backbone=ResNet1012022.12 | 93.4 | — | — | — | — | — | — | — | |
| L2SP (Default)learning rate=0.025, regularization=L2SP2021.07 | 93.29 | — | — | — | — | — | — | — | |
| DCLBase Model=ResNet502021.01 | 93 | — | — | — | — | — | — | — | |
| DCLType=Feature Based, Backbone=ResNet502022.12 | 93 | — | — | — | — | — | — | — | |
| API-NetType=Feature Based, Backbone=ResNet502022.12 | 93 | — | — | — | — | — | — | — | |
| CIN1-Stage=true, Backbone=ResNet-1012020.03 | 92.8 | — | — | — | — | — | — | — | |
| CINType=Feature Based, Backbone=ResNet1012022.12 | 92.8 | — | — | — | — | — | — | — | |
| MoCo V2 Anisotropictraining_mode=SSL, texture_suppression=anisotropic2020.11 | 92.71 | — | — | — | — | — | — | — | |
| Cross-XType=Feature Based, Backbone=ResNet502022.12 | 92.7 | — | — | — | — | — | — | — | |
| CIN1-Stage=true, Backbone=ResNet-502020.03 | 92.6 | — | — | — | — | — | — | — | |
| CINType=Feature Based, Backbone=ResNet502022.12 | 92.6 | — | — | — | — | — | — | — | |
| TASNBase Model=ResNet502021.01 | 92.56 | — | — | — | — | — | — | — | |
| ACNetBase Model=ResNet502021.01 | 92.4 | — | — | — | — | — | — | — | |
| ACNetType=Feature Based, Backbone=ResNet502022.12 | 92.4 | — | — | — | — | — | — | — | |
| DFL-CNN1-Stage=false2020.03 | 92 | — | — | — | — | — | — | — | |
| DFL-CNNType=Feature Based, Backbone=ResNet502022.12 | 92 | — | — | — | — | — | — | — | |
| CBAMBase Model=ResNet502021.01 | 91.91 | — | — | — | — | — | — | — | |
| DFLBase Model=ResNet502021.01 | 91.73 | — | — | — | — | — | — | — | |
| MoCo V2training_mode=SSL2020.11 | 91.57 | — | — | — | — | — | — | — | |
| OursBase Model=VGG192021.01 | 91.55 | — | — | — | — | — | — | — | |
| ACNetType=Feature Based, Backbone=VGG-162022.12 | 91.5 | — | — | — | — | — | — | — | |
| NTSBase Model=ResNet502021.01 | 91.48 | — | — | — | — | — | — | — | |
| NTS1-Stage=false2020.03 | 91.4 | — | — | — | — | — | — | — | |
| NTSNetType=Part Based, Backbone=ResNet502022.12 | 91.4 | — | — | — | — | — | — | — | |
| SENetBase Model=ResNet502021.01 | 91.37 | — | — | — | — | — | — | — | |
| Supervised (Reproduced)training_mode=supervised2020.11 | 90.88 | — | — | — | — | — | — | — | |
| TASNBase Model=VGG192021.01 | 90.83 | — | — | — | — | — | — | — | |
| SaSPA + CutMixFT Strategy=No fine-tuning, Backbone=ResNet50@4482024.06 | 90.79 | — | — | — | — | — | — | — | |
| SaSPA + CutMixBackbone=ResNet50, Resolution=448x448, FT Strategy=No fine-tuning2024.06 | 90.79 | — | — | — | — | — | — | — | |
| SaSPAFT Strategy=No fine-tuning, Backbone=ResNet50@4482024.06 | 90.59 | — | — | — | — | — | — | — | |
| SaSPABackbone=ResNet50, Resolution=448x448, FT Strategy=No fine-tuning2024.06 | 90.59 | — | — | — | — | — | — | — | |
| MA-CNNBase Model=VGG192021.01 | 90.35 | — | — | — | — | — | — | — | |
| CBAMBase Model=VGG192021.01 | 90.32 | — | — | — | — | — | — | — | |
| HBP1-Stage=true2020.03 | 90.3 | — | — | — | — | — | — | — | |
| HBPType=Feature Based, Backbone=VGG-162022.12 | 90.3 | — | — | — | — | — | — | — | |
| Diff-MixFT Strategy=TI+DB, Backbone=ResNet50@4482024.06 | 90.25 | — | — | — | — | — | — | — | |
| Diff-MixBackbone=ResNet50, Resolution=448x448, FT Strategy=Textual Inversion + DreamBooth2024.06 | 90.25 | — | — | — | — | — | — | — | |
| SENetBase Model=VGG192021.01 | 90.12 | — | — | — | — | — | — | — | |
| Diff-Mix + CutMixFT Strategy=TI+DB, Backbone=ResNet50@4482024.06 | 90.01 | — | — | — | — | — | — | — | |
| Diff-Mix + CutMixBackbone=ResNet50, Resolution=448x448, FT Strategy=Textual Inversion + DreamBooth2024.06 | 90.01 | — | — | — | — | — | — | — | |
| MA-CNN1-Stage=false2020.03 | 89.9 | — | — | — | — | — | — | — | |
| MA-CNNType=Part Based, Backbone=VGG-192022.12 | 89.9 | — | — | — | — | — | — | — | |
| MaxEntType=Feature Based, Backbone=DenseNet-1612022.12 | 89.8 | — | — | — | — | — | — | — | |
| iSQRT-COV(8k)1-Stage=true2020.03 | 89.5 | — | — | — | — | — | — | — | |
| CutMixFT Strategy=-, Backbone=ResNet50@4482024.06 | 89.44 | — | — | — | — | — | — | — | |
| CutMixBackbone=ResNet50, Resolution=448x448, FT Strategy=Standard2024.06 | 89.44 | — | — | — | — | — | — | — | |
| MixupBackbone=ResNet50, Resolution=448x448, FT Strategy=Standard2024.06 | 89.41 | — | — | — | — | — | — | — | |
| PCType=Feature Based, Backbone=DenseNet-1612022.12 | 89.2 | — | — | — | — | — | — | — | |
| BaselineBackbone=ResNet50, Resolution=448x448, FT Strategy=Standard2024.06 | 89.09 | — | — | — | — | — | — | — | |
| Real-guidanceBackbone=ResNet50, Resolution=448x448, FT Strategy=No fine-tuning2024.06 | 89.07 | — | — | — | — | — | — | — | |
| ReLICv2Evaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 88.7 | — | — | — | — | — | — | — | |
| Real-filteringBackbone=ResNet50, Resolution=448x448, FT Strategy=No fine-tuning2024.06 | 88.54 | — | — | — | — | — | — | — | |
| Boost-CNN1-Stage=false2020.03 | 88.5 | — | — | — | — | — | — | — | |
| RA-CNNBase Model=VGG192021.01 | 88.2 | — | — | — | — | — | — | — | |
| RA-CNNType=Part Based, Backbone=VGG-192022.12 | 88.2 | — | — | — | — | — | — | — | |
| SimCLREvaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 88.1 | — | — | — | — | — | — | — | |
| BYOLEvaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 88.1 | — | — | — | — | — | — | — | |
| SimCLRtraining_mode=SSL2020.11 | 88.1 | — | — | — | — | — | — | — | |
| BYOLtraining_mode=SSL2020.11 | 88.1 | — | — | — | — | — | — | — | |
| DA-fusionBackbone=ResNet50, Resolution=448x448, FT Strategy=Textual Inversion2024.06 | 87.64 | — | — | — | — | — | — | — | |
| Supervised-INEvaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 86 | — | — | — | — | — | — | — | |
| Random InitEvaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=None2022.01 | 85.9 | — | — | — | — | — | — | — | |
| CGC2021.10 | 85.72 | — | — | — | — | — | — | — | |
| SaSPA + CutMixFT Strategy=No fine-tuning, Backbone=ViT-B/16@3842024.06 | 85.72 | — | — | — | — | — | — | — | |
| SaSPA + CutMixBackbone=ViT-B/16, Resolution=384x384, FT Strategy=No fine-tuning2024.06 | 85.72 | — | — | — | — | — | — | — | |
| SaSPAFT Strategy=No fine-tuning, Backbone=ViT-B/16@3842024.06 | 85.48 | — | — | — | — | — | — | — | |
| SaSPABackbone=ViT-B/16, Resolution=384x384, FT Strategy=No fine-tuning2024.06 | 85.48 | — | — | — | — | — | — | — | |
| MMCLBackbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Setting=Fine-tuning2021.12 | 85.38 | — | — | — | — | — | — | — | |
| Diff-Mix + CutMixFT Strategy=TI+DB, Backbone=ViT-B/16@3842024.06 | 85.12 | — | — | — | — | — | — | — | |
| Diff-Mix + CutMixBackbone=ViT-B/16, Resolution=384x384, FT Strategy=Textual Inversion + DreamBooth2024.06 | 85.12 | — | — | — | — | — | — | — | |
| Diff-MixFT Strategy=TI+DB, Backbone=ViT-B/16@3842024.06 | 84.33 | — | — | — | — | — | — | — | |
| Diff-MixBackbone=ViT-B/16, Resolution=384x384, FT Strategy=Textual Inversion + DreamBooth2024.06 | 84.33 | — | — | — | — | — | — | — | |
| MixupBackbone=ViT-B/16, Resolution=384x384, FT Strategy=Standard2024.06 | 84.31 | — | — | — | — | — | — | — | |
| BilinearType=Feature Based, Backbone=VGG-162022.12 | 84.1 | — | — | — | — | — | — | — | |
| Baseline2021.10 | 83.65 | — | — | — | — | — | — | — | |
| SupervisedBackbone=ResNet-50, Evaluation Setting=Fine-tuning2021.12 | 83.5 | — | — | — | — | — | — | — | |
| CutMixFT Strategy=-, Backbone=ViT-B/16@3842024.06 | 83.5 | — | — | — | — | — | — | — | |
| BaselineBackbone=ViT-B/16, Resolution=384x384, FT Strategy=Standard2024.06 | 83.5 | — | — | — | — | — | — | — | |
| CutMixBackbone=ViT-B/16, Resolution=384x384, FT Strategy=Standard2024.06 | 83.5 | — | — | — | — | — | — | — | |
| Real-guidanceBackbone=ViT-B/16, Resolution=384x384, FT Strategy=No fine-tuning2024.06 | 83.17 | — | — | — | — | — | — | — | |
| Real-filteringBackbone=ViT-B/16, Resolution=384x384, FT Strategy=No fine-tuning2024.06 | 83.07 | — | — | — | — | — | — | — | |
| MOCHIBackbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Setting=Fine-tuning, Public Checkpoint=true2021.12 | 83.03 | — | — | — | — | — | — | — | |
| MoCo-v2Backbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Setting=Fine-tuning, Public Checkpoint=true2021.12 | 82.46 | — | — | — | — | — | — | — | |
| DA-fusionBackbone=ViT-B/16, Resolution=384x384, FT Strategy=Textual Inversion2024.06 | 81.88 | — | — | — | — | — | — | — | |
| TiF learner w/o cBackbone=Diffusion Model (SD 2.0), N (number of shots)=162024.03 | 80.4 | — | — | — | — | — | — | — | |
| TiF learnerBackbone=Diffusion Model (SD 2.0), N (number of shots)=162024.03 | 79.9 | — | — | — | — | — | — | — | |
| PCL-v2Backbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Setting=Fine-tuning2021.12 | 79.37 | — | — | — | — | — | — | — | |
| MoCoBackbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Setting=Fine-tuning2021.12 | 75.61 | — | — | — | — | — | — | — | |
| PCL-v1Backbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Setting=Fine-tuning2021.12 | 74.97 | — | — | — | — | — | — | — |