https://docs.google.com/spreadsheets/d/15H_7U2f410hnoP8cLz57I7Y1M4aaXn3F4iIskYBrNwM/edit?usp=sharing
AlexNet | Imagenet classification with deep convolutional neural networks | 2012-NIPS | paper | |
ZF | Visualizing and Understanding Convolutional Networks | 2014-ECCV | arxiv | |
OverFeat | OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks | 2014-ICLR | arxiv | |
VGG | Very deep convolutional networks for large-scale image recognition | 2015-ICLR | arxiv | |
Inception v1 | Going Deeper with Convolutions | 2015-CVPR | arxiv | |
Inception v2 | Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift | 2015-ICML | arxiv | |
Inception v3 | Rethinking the Inception Architecture for Computer Vision | 2016-CVPR | arxiv | |
Inception v4 | Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning | 2017-AAAI | arxiv | |
ResNet,ORU | deep residual learning for image recognition | 2016-CVPR | arxiv | code |
ResNet,PRU | Identity Mappings in Deep Residual Networks | 2016-ECCV | arxiv | code |
ResNeXt | Aggregated Residual Transformations for Deep Neural Networks | 2017-CVPR | arxiv | code |
DenseNet | Densely Connected Convolutional Networks | 2017-CVPR | arxiv | code |
ILSVRC2012 | ||||||||
---|---|---|---|---|---|---|---|---|
params | FLOPs | size | augmentation | model | crops | top1 | top5 | |
DenseNet-121 | 1 | 1 | 25.02 | 7.71 | ||||
DenseNet-169 | 1 | 1 | 23.8 | 6.85 | ||||
DenseNet-201 | 1 | 1 | 22.58 | 6.34 | ||||
DenseNet-264 | 1 | 1 | 22.15 | 6.12 | ||||
DenseNet-121 | 1 | 10 | 23.61 | 6.66 | ||||
DenseNet-169 | 1 | 10 | 22.08 | 5.92 | ||||
DenseNet-201 | 1 | 10 | 21.46 | 5.54 | ||||
DenseNet-264 | 1 | 10 | 20.8 | 5.29 | ||||
ResNeXt-50 | 224 | 1 | 1 | 24.4 | 6.6 | |||
ResNeXt-101 | 224 | 1 | 1 | 22.2 | 5.7 | |||
Inception-ResNet-v4 | 1 | 1 | 21.3 | 5.5 | ||||
Inception-v4 | 1 | 1 | 20 | 5 | ||||
Inception-ResNet-v2 | 1 | 1 | 19.9 | 4.9 | ||||
Inception-ResNet-v1 | 1 | 12 | 19.8 | 4.6 | ||||
Inception-v4 | 1 | 12 | 18.7 | 4.2 | ||||
Inception-ResNet-v2 | 1 | 12 | 18.7 | 4.1 | ||||
Inception-ResNet-v1 | 1 | 144 | 18.8 | 4.3 | ||||
Inception-v4 | 1 | 144 | 17.7 | 3.8 | ||||
Inception-ResNet-v2 | 1 | 144 | 17.8 | 3.7 | ||||
Inception-v4 +3× Inception-ResNet-v2 | 4 | 144 | 16.5 | 3.1 | ||||
Resnet-50,ORU | 1 | 20.74 | 5.25 | |||||
Resnet-101,ORU | 1 | 19.87 | 4.6 | |||||
Resnet-152,ORU | 11.3b | 1 | 19.38 | 4.49 | ||||
ResNet-152, ORU | 224/224 | scale | 1 | 23 | 6.7 | |||
ResNet-152, ORU | 224/320 | scale | 1 | 21.3 | 5.5 | |||
ResNet-152, PRU | 224/320 | scale | 1 | 21.1 | 5.5 | |||
ResNet-200, ORU | 224/320 | scale | 1 | 21.8 | 6 | |||
ResNet-200, PRU | 224/320 | scale | 1 | 20.7 | 5.3 | |||
ResNet-200, PRU | 224/320 | scale,asp ratio | 1 | 20.1 | 4.8 | |||
Inception-v3 | 299 | 1 | 12 | 19.47 | 4.48 | |||
Inception-v3 | 299 | 1 | 144 | 18.77 | 4.2 | |||
Inception-v3 | 299 | 4 | 144 | 17.2 | 3.58 | |||
Inception-v2 | 1 | 1 | 25.2 | 7.82 | ||||
Inception-v2 | 1 | 144 | 21.99 | 5.82 | ||||
Inception-v2 | 6 | 144 | 20.1 | 4.9 | ||||
Inception-v1 | 224 | 1 | 1 | 10.07 | ||||
Inception-v1 | 224 | 1 | 10 | 9.15 | ||||
Inception-v1 | 224 | 1 | 144 | 7.89 | ||||
Inception-v1 | 224 | 7 | 1 | 8.09 | ||||
Inception-v1 | 224 | 7 | 10 | 7.62 | ||||
Inception-v1 | 224 | 7 | 144 | 6.67 | ||||
VGG-16 | 256 | 27 | 8.8 | |||||
VGG-16 | 384 | 26.8 | 8.7 | |||||
VGG-16 | [256;512]/384 | 25.6 | 8.1 | |||||
VGG-16 | 256/[224;256;288] | 26.6 | 8.6 | |||||
VGG-16 | 384/[352;384;416] | 26.5 | 8.6 | |||||
VGG-16 | [256;512]/[256;384;512] | 26.5 | 7.5 | |||||
VGG-16 | 24.6 | 7.5 | ||||||
VGG-16 | 24.4 | 7.2 | ||||||
VGG-19 | 256 | 27.3 | 9 | |||||
VGG-19 | 384 | 26.9 | 8.7 | |||||
VGG-19 | [256;512]/384 | 25.5 | 8 | |||||
VGG-19 | 256/[224;256;288] | 26.9 | 8.7 | |||||
VGG-19 | 384/[352;384;416] | 26.7 | 8.6 | |||||
VGG-19 | [256;512]/[256;384;512] | 24.8 | 7.5 | |||||
VGG-19 | 24.6 | 7.4 | ||||||
VGG-19 | 24.4 | 7.1 |