/cifar10-pytorch-cnn

Simple implementation of a residually connected convolutional neural network in PyTorch

Primary LanguagePythonMIT LicenseMIT

CIFAR 10 CNN in PyTorch

Performance

Test Accuracy: 87.31% Train Accuracy: 99.92%

How to run

python main.py

Info

The model uses residual connections between the start and end of each Convolutional block. However, partial residual connections have not been implemented for blocks with different input and output size.

Required Packages

  • pytorch
  • torchvision
  • tensorboard
  • scikit-learn
  • numpy
  • tqdm