/pytorch-classifiers

PyTorch implementation of classic classifiers networks.

Primary LanguagePython

PyTorch Classifiers

PyTorch implementation of the classic classifiers network.

Implemented Classifiers

  • LeNet
  • AlexNet
  • VGG16
  • GoogLeNet (Inception-v1)
  • ResNet-18, ResNet-34, ResNet-50, ResNet-101 and ResNet-152
  • Inception-v3, Inception-v4
  • DenseNet
  • SENet
  • EfficientNet

Results

Accuracy on MNIST

Model Parameters Batch Size Iterations Accuracy
LeNet-5 0.0371M 64 10k 97.8404%
AlexNet 58.2991M 64 10k 99.0048%
VGG16 134.3004M 16 2k 97.6600%

Accuracy on CIFAR-10

Model Parameters Batch Size Iterations Accuracy
LeNet-5 0.0371M 64 10k 52.1994%
AlexNet 58.3223M 64 10k 80.5434%
VGG16 134.3015M 16 2k 54.5800%

Accuracy on CIFAR-100

Model Parameters Batch Size Iterations Accuracy
LeNet-5 0.0448M 64 10k 15.8340%
AlexNet 58.6910M 64 10k 48.8256%
VGG16 134.6702M 16 10k 43.1100%
GoogLeNet 10.6183M 32 10k 44.8582%
GoogLeNet 10.6183M 32 20k 53.0152%
GoogLeNet 10.6183M 64 10k 50.1393%
ResNet-18 11.2278M 64 10k 53.4833%

Architectures

LeNet

arch Image Credit

AlexNet

arch Image Credit

VGG16

arch Image Credit

GoogLeNet

arch Image Credit

ResNet

arch Image Credit

ResNet-18

arch Image Credit

ResNet-34

arch Image Credit

Others

This repository is for practice purposes only.