/CIFAR-10-CNN

Convolutional neural network to classify the CIFAR10 dataset

Primary LanguageJupyter Notebook

CIFAR-10 Convolutional Neural Network

Implementation of two models in Keras for the classification of the CIFAR-10 dataset.

First model

It's a simple model with three Convolution+Max Pooling+Dropuout layers and two fully-connected layers. It has a validation accuracy of 78%.

Second model

The second model uses augmentation and three different types of normalization:

  • Batch normalization
  • Kernel regularization
  • Dropout

I also use a learning rate scheduler to change the lr in the middle of the training

README WIP