Skip to content

Latest commit

 

History

History
62 lines (54 loc) · 2.13 KB

README.md

File metadata and controls

62 lines (54 loc) · 2.13 KB
Raw

Train and test CIFAR10 with tensorflow

introduction

various nets implement train and test in cifar10 dataset with tensorflow,the nets include VGG,Resnet,Resnext,mobilenet,SENet,xception and so on

Prepare data,Testing,Training

  1. Clone the cifar-tensorflow repository
 git https://github.com/yxlijun/cifar-tensorflow
  1. Prepare data,run folloing code,generate cifar10_data directory
python tools/cifar10_download_and_extract.py
  1. training
 python train.py --net vgg16
  1. testing
python test.py --net vgg16

####Accuracy

Model Acc.
VGG11 91.35%
VGG13 93.02%
VGG16 93.62%
VGG19 93.75%
Resnet20 94.43%
Resnet32 94.73%
Resnet44 94.82%
Resnet56 95.04%
Xception 95.11%
MobileNet 95.16%
DensetNet40_12 94.24%
DenseNet100_12 95.21%
DenseNet100_24 95.21%
DenseNet100_24 95.21%
ResNext50 95.21%
ResNext101 95.21%
SqueezeNetA 95.21%
SqueezeNetB 95.21%
SE_Resnet_50 95.21%
SE_Resnet_101 95.21%

Net implement

  • VGG
  • ResNet
  • DenseNet
  • mobileNet
  • ResNext
  • Xception
  • SeNet
  • SqueenzeNet