| Rank | Model |
Percentage correct | Top-1 Accuracy | Accuracy | Parameters | Top 1 Accuracy | F1 | Cross Entropy Loss |
Extra Training Data |
Paper | Code | Year |
| 1 |
ViT-H/14 |
99.5 | – | – | – | – | – | – |
✓ |
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
|
huggingface/transformers · labmlai/annotated_deep_learning_paper_implementations · rwightman/pytorch-image-models
· +155 |
2020 |
| 1 |
DINOv2 (ViT-g/14, frozen model, linear eval) |
99.5 | – | – | – | – | – | – |
✓ |
DINOv2: Learning Robust Visual Features without Supervision
|
huggingface/transformers · facebookresearch/dinov2 · roboflow/rf-detr
· +23 |
2023 |
| 3 |
µ2Net (ViT-L/16) |
99.49 | – | – | – | – | – | – |
✓ |
An Evolutionary Approach to Dynamic Introduction of Tasks in Large-scale Multitask Learning Systems
|
google-research/google-research |
2022 |
| 4 |
ViT-L/16 |
99.42 | – | – | – | – | – | – |
✓ |
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
|
huggingface/transformers · labmlai/annotated_deep_learning_paper_implementations · rwightman/pytorch-image-models
· +155 |
2020 |
| 5 |
CaiT-M-36 U 224 |
99.4 | – | – | – | – | – | – |
✓ |
Going deeper with Image Transformers
|
rwightman/pytorch-image-models · lucidrains/vit-pytorch · facebookresearch/deit
· +18 |
2021 |
| 6 |
CvT-W24 |
99.39 | – | – | – | – | – | – |
✓ |
CvT: Introducing Convolutions to Vision Transformers
|
huggingface/transformers · BR-IDL/PaddleViT · microsoft/CvT
· +13 |
2021 |
| 7 |
BiT-L (ResNet) |
99.37 | – | – | – | – | – | – |
✓ |
Big Transfer (BiT): General Visual Representation Learning
|
google-research/big_transfer · sayakpaul/FunMatch-Distillation · bethgelab/InDomainGeneralizationBenchmark
· +6 |
2019 |
| 8 |
RDNet-L (224 res, IN-1K pretrained) |
99.31 | – | – | – | – | – | – |
✓ |
DenseNets Reloaded: Paradigm Shift Beyond ResNets and ViTs
|
huggingface/pytorch-image-models · naver-ai/rdnet · birder/birder |
2024 |
| 8 |
RDNet-B (224 res, IN-1K pretrained) |
99.31 | – | – | – | – | – | – |
✓ |
DenseNets Reloaded: Paradigm Shift Beyond ResNets and ViTs
|
huggingface/pytorch-image-models · naver-ai/rdnet · birder/birder |
2024 |
| 10 |
ViT-B (attn fine-tune) |
99.3 | – | – | – | – | – | – |
✓ |
Three things everyone should know about Vision Transformers
|
rwightman/pytorch-image-models · lucidrains/vit-pytorch · facebookresearch/deit
· +5 |
2022 |
| 11 |
Heinsen Routing + BEiT-large 16 224 |
99.2 | – | – | – | – | – | – |
✓ |
An Algorithm for Routing Vectors in Sequences
|
glassroom/heinsen_routing |
2022 |
| 12 |
ViT-B/16 (PUGD) |
99.13 | – | – | – | – | – | – |
✓ |
Perturbated Gradients Updating within Unit Space for Deep Learning
|
hanktseng131415go/pugd |
2021 |
| 13 |
Astroformer |
99.12 | 99.12 | – | – | – | – | – |
✓ |
Astroformer: More Data Might not be all you need for Classification
|
Rishit-dagli/Astroformer |
2023 |
| 14 |
DeiT-B |
99.1 | – | – | – | – | – | – |
✓ |
Training data-efficient image transformers & distillation through attention
|
huggingface/transformers · rwightman/pytorch-image-models · PaddlePaddle/PaddleClas
· +37 |
2020 |
| 14 |
TNT-B |
99.1 | – | – | – | – | – | – |
✓ |
Transformer in Transformer
|
rwightman/pytorch-image-models · PaddlePaddle/PaddleClas · huawei-noah/CV-backbones
· +9 |
2021 |
| 14 |
CeiT-S (384 finetune resolution) |
99.1 | – | – | – | – | – | – |
✓ |
Incorporating Convolution Designs into Visual Transformers
|
rishikksh20/CeiT-pytorch · coeusguo/ceit · mindspore-courses/External-Attention-MindSpore |
2021 |
| 14 |
EfficientNetV2-L |
99.1 | – | – | – | – | – | – |
✓ |
EfficientNetV2: Smaller Models and Faster Training
|
rwightman/pytorch-image-models · pytorch/vision · lukemelas/EfficientNet-PyTorch
· +23 |
2021 |
| 14 |
AutoFormer-S | 384 |
99.1 | – | – | – | – | – | – |
✓ |
AutoFormer: Searching Transformers for Visual Recognition
|
microsoft/AutoML · microsoft/cream |
2021 |
| 19 |
VIT-L/16 (Spinal FC, Background) |
99.05 | – | – | – | – | – | – |
|
Reduction of Class Activation Uncertainty with Background Information
|
dipuk0506/SpinalNet · dipuk0506/uq |
2023 |
| 20 |
LaNet |
99.03 | – | – | – | – | – | – |
|
Sample-Efficient Neural Architecture Search by Learning Action Space for Monte Carlo Tree Search
|
facebookresearch/LaMCTS |
2019 |
| 21 |
GPIPE + transfer learning |
99 | – | – | – | – | – | – |
✓ |
GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism
|
tensorflow/lingvo · qubvel/efficientnet · KakaoBrain/torchgpipe
· +10 |
2018 |
| 21 |
TResNet-XL |
99 | – | – | – | – | – | – |
|
TResNet: High Performance GPU-Dedicated Architecture
|
rwightman/pytorch-image-models · mrT23/TResNet · Alibaba-MIIL/TResNet |
2020 |
| 21 |
CeiT-S |
99 | – | – | – | – | – | – |
|
Incorporating Convolution Designs into Visual Transformers
|
rishikksh20/CeiT-pytorch · coeusguo/ceit · mindspore-courses/External-Attention-MindSpore |
2021 |
| 21 |
EfficientNetV2-M |
99.0 | – | – | – | – | – | – |
|
EfficientNetV2: Smaller Models and Faster Training
|
rwightman/pytorch-image-models · pytorch/vision · lukemelas/EfficientNet-PyTorch
· +23 |
2021 |
| 21 |
GFNet-H-B |
99.0 | – | – | – | – | – | – |
✓ |
Global Filter Networks for Image Classification
|
raoyongming/GFNet · liuruiyang98/Jittor-MLP · mindspore-courses/External-Attention-MindSpore
· +1 |
2021 |
| 26 |
BiT-M (ResNet) |
98.91 | – | – | – | – | – | – |
|
Big Transfer (BiT): General Visual Representation Learning
|
google-research/big_transfer · sayakpaul/FunMatch-Distillation · bethgelab/InDomainGeneralizationBenchmark
· +6 |
2019 |
| 27 |
EfficientNet-B7 |
98.9 | – | – | – | – | – | – |
|
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
|
ultralytics/yolov5 · rwightman/pytorch-image-models · open-mmlab/mmdetection
· +141 |
2019 |
| 28 |
RDNet-T (224 res, IN-1K pretrained) |
98.88 | – | – | – | – | – | – |
|
DenseNets Reloaded: Paradigm Shift Beyond ResNets and ViTs
|
huggingface/pytorch-image-models · naver-ai/rdnet · birder/birder |
2024 |
| 29 |
ASF-former-B |
98.8% | – | – | – | – | – | – |
|
Adaptive Split-Fusion Transformer
|
szx503045266/asf-former |
2022 |
| 30 |
PyramidNet-272, S=4 |
98.71 | – | – | – | – | – | – |
|
Towards Better Accuracy-efficiency Trade-offs: Divide and Co-training
|
freeformrobotics/divide-and-co-training · mzhaoshuai/Divide-and-Co-training |
2020 |
| 31 |
EfficientNetV2-S |
98.7 | – | – | – | – | – | – |
|
EfficientNetV2: Smaller Models and Faster Training
|
rwightman/pytorch-image-models · pytorch/vision · lukemelas/EfficientNet-PyTorch
· +23 |
2021 |
| 31 |
ASF-former-S |
98.7 | – | – | – | – | – | – |
|
Adaptive Split-Fusion Transformer
|
szx503045266/asf-former |
2022 |
| 33 |
PyramidNet-272 (ASAM) |
98.68 | – | – | – | – | – | – |
|
ASAM: Adaptive Sharpness-Aware Minimization for Scale-Invariant Learning of Deep Neural Networks
|
davda54/sam · borealisai/perturbed-forgetting |
2021 |
| 34 |
PyramidNet + ShakeDrop + Fast AA + FMix |
98.64 | – | – | – | – | – | – |
|
FMix: Enhancing Mixed Sample Data Augmentation
|
PaddlePaddle/PaddleClas · Westlake-AI/openmixup · ecs-vlc/FMix
· +2 |
2020 |
| 35 |
ViT-B/16- SAM |
98.6 | – | – | – | – | – | – |
|
When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations
|
google-research/vision_transformer · ttt496/VisionTransformer |
2021 |
| 35 |
ConvMLP-M |
98.6 | – | – | – | – | – | – |
✓ |
ConvMLP: Hierarchical Convolutional MLPs for Vision
|
BR-IDL/PaddleViT · shinya7y/UniverseNet · liuruiyang98/Jittor-MLP
· +1 |
2021 |
| 35 |
ConvMLP-L |
98.6 | – | – | – | – | – | – |
✓ |
ConvMLP: Hierarchical Convolutional MLPs for Vision
|
BR-IDL/PaddleViT · shinya7y/UniverseNet · liuruiyang98/Jittor-MLP
· +1 |
2021 |
| 38 |
DVT (T2T-ViT-24) |
98.53 | – | – | – | – | – | – |
|
Not All Images are Worth 16x16 Words: Dynamic Transformers for Efficient Image Recognition
|
blackfeather-wang/Dynamic-Vision-Transformer · blackfeather-wang/dynamic-vision-transformer-mindspore |
2021 |
| 39 |
E2E-3M |
98.52 | – | – | – | – | – | – |
|
Rethinking Recurrent Neural Networks and Other Improvements for Image Classification
|
leonlha/e2e-3m |
2020 |
| 40 |
CeiT-T |
98.5 | – | – | – | – | – | – |
|
Incorporating Convolution Designs into Visual Transformers
|
rishikksh20/CeiT-pytorch · coeusguo/ceit · mindspore-courses/External-Attention-MindSpore |
2021 |
| 41 |
NAT-M4 |
98.4 | 98.4 | – | 6.9M | – | – | – |
✓ |
Neural Architecture Transfer
|
human-analysis/neural-architecture-transfer · awesomelemon/encas |
2020 |
| 42 |
WRN-40-10, S=4 |
98.38 | – | – | – | – | – | – |
|
Towards Better Accuracy-efficiency Trade-offs: Divide and Co-training
|
freeformrobotics/divide-and-co-training · mzhaoshuai/Divide-and-Co-training |
2020 |
| 43 |
WRN-28-10, S=4 |
98.32 | – | – | – | – | – | – |
|
Towards Better Accuracy-efficiency Trade-offs: Divide and Co-training
|
freeformrobotics/divide-and-co-training · mzhaoshuai/Divide-and-Co-training |
2020 |
| 44 |
Shake-Shake 26 2x96d, S=4 |
98.31 | – | – | – | – | – | – |
|
Towards Better Accuracy-efficiency Trade-offs: Divide and Co-training
|
freeformrobotics/divide-and-co-training · mzhaoshuai/Divide-and-Co-training |
2020 |
| 44 |
Dynamics 2 |
98.31 | – | – | – | – | – | – |
|
PSO-Convolutional Neural Networks with Heterogeneous Learning Rate
|
leonlha/pso-convnet-dynamics |
2022 |
| 46 |
PyramidNet+ShakeDrop (Fast AA) |
98.3 | – | – | – | – | – | – |
|
Fast AutoAugment
|
kakaobrain/fast-autoaugment · ildoonet/pytorch-randaugment · philip-bachman/amdim-public
· +8 |
2019 |
| 46 |
ResNet50 (A1) |
98.3 | – | – | – | – | – | – |
|
ResNet strikes back: An improved training procedure in timm
|
rwightman/pytorch-image-models · open-mmlab/mmdetection · open-mmlab/mmclassification
· +11 |
2021 |
| 48 |
NoisyDARTS-A-t |
98.28 | – | – | – | – | – | – |
|
Noisy Differentiable Architecture Search
|
xiaomi-automl/NoisyDARTS |
2020 |
| 49 |
NAT-M3 |
98.2 | 98.2 | – | 6.2M | – | – | – |
✓ |
Neural Architecture Transfer
|
human-analysis/neural-architecture-transfer · awesomelemon/encas |
2020 |
| 49 |
LeViT-192 |
98.2 | – | – | – | – | – | – |
|
LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference
|
huggingface/transformers · rwightman/pytorch-image-models · leondgarse/keras_cv_attention_models
· +9 |
2021 |
| 49 |
ResNet-152-SAM |
98.2 | – | – | – | – | – | – |
|
When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations
|
google-research/vision_transformer · ttt496/VisionTransformer |
2021 |
| 49 |
ViT-S/16- SAM |
98.2 | – | – | – | – | – | – |
|
When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations
|
google-research/vision_transformer · ttt496/VisionTransformer |
2021 |
| 49 |
Bamboo (ViT-B/16) |
98.2 | – | – | – | – | – | – |
|
Bamboo: Building Mega-Scale Vision Dataset Continually with Human-Machine Synergy
|
zhangyuanhan-ai/bamboo · davidzhangyuanhan/bamboo |
2022 |
| 49 |
DE ELBo (ViT-B/16) |
98.2 | – | – | – | – | – | – |
|
Learning Hyperparameters via a Data-Emphasized Variational Objective
|
tufts-ml/data-emphasized-ELBo |
2025 |
| 55 |
LeViT-256 |
98.1 | – | – | – | – | – | – |
✓ |
LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference
|
huggingface/transformers · rwightman/pytorch-image-models · leondgarse/keras_cv_attention_models
· +9 |
2021 |
| 56 |
PyramidNet + AA (AMP) |
98.02 | – | – | – | – | – | – |
|
Regularizing Neural Networks via Adversarial Model Perturbation
|
hiyouga/AMP-Regularizer |
2020 |
| 57 |
EnAET |
98.01 | – | – | – | – | – | – |
|
EnAET: A Self-Trained framework for Semi-Supervised and Supervised Learning with Ensemble Transformations
|
maple-research-lab/EnAET · wang3702/EnAET |
2019 |
| 58 |
MUXNet-m |
98.0 | 98.0 | – | – | – | – | – |
|
MUXConv: Information Multiplexing in Convolutional Neural Networks
|
human-analysis/MUXConv |
2020 |
| 58 |
LeViT-384 |
98 | – | – | – | – | – | – |
|
LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference
|
huggingface/transformers · rwightman/pytorch-image-models · leondgarse/keras_cv_attention_models
· +9 |
2021 |
| 58 |
CCT-7/3x1* |
98 | – | – | – | – | – | – |
|
Escaping the Big Data Paradigm with Compact Transformers
|
keras-team/keras-io · SHI-Labs/Compact-Transformers · brohrer/sharpened-cosine-similarity
· +6 |
2021 |
| 58 |
ConvMLP-S |
98 | – | – | – | – | – | – |
✓ |
ConvMLP: Hierarchical Convolutional MLPs for Vision
|
BR-IDL/PaddleViT · shinya7y/UniverseNet · liuruiyang98/Jittor-MLP
· +1 |
2021 |
| 62 |
Proxyless-G + c/o |
97.92 | – | – | – | – | – | – |
|
ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware
|
osmr/imgclsmob · mit-han-lab/once-for-all · mit-han-lab/ProxylessNAS
· +20 |
2018 |
| 63 |
NAT-M2 |
97.9 | 97.9 | – | 4.6M | – | – | – |
✓ |
Neural Architecture Transfer
|
human-analysis/neural-architecture-transfer · awesomelemon/encas |
2020 |
| 63 |
WRN-28-10+AutoDropout+RandAugment |
97.9 | – | – | – | – | – | – |
|
AutoDropout: Learning Dropout Patterns to Regularize Deep Networks
|
google-research/google-research |
2021 |
| 65 |
SENet + ShakeShake + Cutout |
97.88 | – | – | – | – | – | – |
|
Squeeze-and-Excitation Networks
|
PaddlePaddle/PaddleOCR · xmu-xiaoma666/External-Attention-pytorch · open-mmlab/mmpose
· +82 |
2017 |
| 66 |
HCGNet-A3 |
97.86 | – | – | – | – | – | – |
|
Gated Convolutional Networks with Hybrid Connectivity for Image Classification
|
winycg/HCGNet |
2019 |
| 67 |
Wide-ResNet-28-10 |
97.85 | – | – | – | – | – | – |
|
Automatic Data Augmentation via Invariance-Constrained Learning
|
ihounie/daug |
2022 |
| 68 |
ResNeXt-50 (AutoMix) |
97.84 | – | – | – | – | – | – |
|
AutoMix: Unveiling the Power of Mixup for Stronger Classifiers
|
Westlake-AI/openmixup · zeyuanyin/tiny-imagenet · Westlake-AI/AutoMix |
2021 |
| 69 |
ResNet-152x4-AGC (ImageNet-21K) |
97.82 | – | – | – | – | – | – |
|
Effect of Pre-Training Scale on Intra- and Inter-Domain Full and Few-Shot Transfer Learning for Natural and Medical X-Ray Chest Images
|
SLAMPAI/large-scale-pretraining-transfer |
2021 |
| 70 |
Mixer-B/16- SAM |
97.8 | – | – | – | – | – | – |
|
When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations
|
google-research/vision_transformer · ttt496/VisionTransformer |
2021 |
| 71 |
CCT-7/3x1+VTM |
97.78 | – | – | – | – | – | – |
|
TokenMixup: Efficient Attention-guided Token-level Data Augmentation for Transformers
|
mlvlab/tokenmixup |
2022 |
| 72 |
WRN-28-10 |
97.73 | – | – | – | – | – | – |
|
MixMo: Mixing Multiple Inputs for Multiple Outputs via Deep Subnetworks
|
alexrame/mixmo-pytorch |
2021 |
| 73 |
HCGNet-A2 |
97.71 | – | – | – | – | – | – |
|
Gated Convolutional Networks with Hybrid Connectivity for Image Classification
|
winycg/HCGNet |
2019 |
| 74 |
WRN + fixup init + mixup + cutout |
97.7 | – | – | – | – | – | – |
|
Fixup Initialization: Residual Learning Without Normalization
|
hongyi-zhang/Fixup · bzhangGo/zero · Zelgunn/CustomKerasLayers
· +7 |
2019 |
| 75 |
NoisyDARTS-a |
97.61 | – | – | – | – | – | – |
|
Noisy Differentiable Architecture Search
|
xiaomi-automl/NoisyDARTS |
2020 |
| 75 |
TransBoost-ResNet50 |
97.61 | – | – | – | – | – | – |
|
TransBoost: Improving the Best ImageNet Performance using Deep Transduction
|
omerb01/transboost |
2022 |
| 77 |
LeViT-128 |
97.6 | – | – | – | – | – | – |
|
LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference
|
huggingface/transformers · rwightman/pytorch-image-models · leondgarse/keras_cv_attention_models
· +9 |
2021 |
| 78 |
DenseNet-BC-190 + batchboost |
97.54 | – | – | – | – | – | – |
|
batchboost: regularization for stabilizing training with resistance to underfitting & overfitting
|
maciejczyzewski/batchboost |
2020 |
| 79 |
LeViT-128S |
97.5 | – | – | – | – | – | – |
✓ |
LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference
|
huggingface/transformers · rwightman/pytorch-image-models · leondgarse/keras_cv_attention_models
· +9 |
2021 |
| 80 |
Shared WRN |
97.47 | – | – | – | – | – | – |
|
Learning Implicitly Recurrent CNNs Through Parameter Sharing
|
lolemacs/soft-sharing |
2019 |
| 81 |
Manifold Mixup WRN 28-10 |
97.45 | – | – | – | – | – | – |
|
Manifold Mixup: Better Representations by Interpolating Hidden States
|
Westlake-AI/openmixup · vikasverma1077/manifold_mixup · jankrepl/mildlyoverfitted
· +9 |
2018 |
| 81 |
WRN 28-14 |
97.45 | – | – | – | – | – | – |
|
Neural networks with late-phase weights
|
google/uncertainty-baselines · seijin-kobayashi/late-phase-weights |
2020 |
| 83 |
SparseSwin |
97.43 | – | – | – | – | – | – |
|
SparseSwin: Swin Transformer with Sparse Transformer Block
|
krisnapinasthika/sparseswin |
2023 |
| 84 |
WRN-28-10 with reSGHMC |
97.42 | – | – | – | – | – | – |
|
Non-convex Learning via Replica Exchange Stochastic Gradient MCMC
|
gaoliyao/Replica_Exchange_Stochastic_Gradient_MCMC · WayneDW/Variance_Reduced_Replica_Exchange_SGMCMC |
2020 |
| 85 |
NAT-M1 |
97.4 | 97.4 | – | 4.3M | – | – | – |
✓ |
Neural Architecture Transfer
|
human-analysis/neural-architecture-transfer · awesomelemon/encas |
2020 |
| 85 |
ResNet-50-SAM |
97.4 | – | – | – | – | – | – |
✓ |
When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations
|
google-research/vision_transformer · ttt496/VisionTransformer |
2021 |
| 87 |
DenseNet-BC-190 + Mixup |
97.3 | – | – | – | – | – | – |
✓ |
mixup: Beyond Empirical Risk Minimization
|
rwightman/pytorch-image-models · pytorch/vision · PaddlePaddle/PaddleDetection
· +68 |
2017 |
| 87 |
kNN-CLIP |
97.3 | – | – | – | – | – | – |
✓ |
Revisiting a kNN-based Image Classification System with High-capacity Storage
|
|
2022 |
| 89 |
WaveMixLite-144/7 |
97.29 | – | – | – | – | – | – |
|
WaveMix: A Resource-efficient Neural Network for Image Analysis
|
pranavphoenix/WaveMix |
2022 |
| 90 |
Transformer local-attention (NesT-B) |
97.2 | – | – | – | – | – | – |
|
Nested Hierarchical Transformer: Towards Accurate, Data-Efficient and Interpretable Visual Understanding
|
rwightman/pytorch-image-models · google-research/nested-transformer · ttt496/vit-pytorch
· +3 |
2021 |
| 91 |
ShakeShake-2x64d + SWA |
97.12 | – | – | – | – | – | – |
|
Averaging Weights Leads to Wider Optima and Better Generalization
|
timgaripov/swa · wjmaddox/swa_gaussian · ENSTA-U2IS-AI/torch-uncertainty
· +14 |
2018 |
| 91 |
PyramidNet-200 + CutMix |
97.12 | – | – | – | – | – | – |
|
CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features
|
rwightman/pytorch-image-models · pytorch/vision · kornia/kornia
· +27 |
2019 |
| 93 |
Wide-ResNet-40-2 |
97.05 | – | – | – | – | – | – |
|
Automatic Data Augmentation via Invariance-Constrained Learning
|
ihounie/daug |
2022 |
| 94 |
ORN |
97.02 | – | – | – | – | – | – |
|
Oriented Response Networks
|
ZhouYanzhao/ORN |
2017 |
| 95 |
WRN-16-8 with reSGHMC |
96.87 | – | – | – | – | – | – |
|
Non-convex Learning via Replica Exchange Stochastic Gradient MCMC
|
gaoliyao/Replica_Exchange_Stochastic_Gradient_MCMC · WayneDW/Variance_Reduced_Replica_Exchange_SGMCMC |
2020 |
| 95 |
ResNet_XnIDR |
96.87 | – | – | – | – | – | – |
|
XnODR and XnIDR: Two Accurate and Fast Fully Connected Layers For Convolutional Neural Networks
|
jiansfoggy/CODE-SHOW |
2021 |
| 97 |
HCGNet-A1 |
96.85 | – | – | – | – | – | – |
|
Gated Convolutional Networks with Hybrid Connectivity for Image Classification
|
winycg/HCGNet |
2019 |
| 98 |
WRN 28-10 |
96.81 | – | – | – | – | – | – |
|
Neural networks with late-phase weights
|
google/uncertainty-baselines · seijin-kobayashi/late-phase-weights |
2020 |
| 99 |
AutoDropout |
96.8 | – | – | – | – | – | – |
|
AutoDropout: Learning Dropout Patterns to Regularize Deep Networks
|
google-research/google-research |
2021 |
| 100 |
WRN-28-10 + SWA |
96.79 | – | – | – | – | – | – |
|
Averaging Weights Leads to Wider Optima and Better Generalization
|
timgaripov/swa · wjmaddox/swa_gaussian · ENSTA-U2IS-AI/torch-uncertainty
· +14 |
2018 |