This is a Keras model based on VGG16 architecture for CIFAR-10 and CIFAR-100. it can be used either with pretrained weights file or trained from scratch.

This package contains 2 classes one for each datasets, the architecture is based on the VGG-16 [1] with adaptation to CIFAR datasets based on [2]. By running the py files you can get a sample of a trining and estimation of validation error.

The CIFAR-10 reaches a validation accuracy of 93.56% CIFAR-100 reaches validation accuracy of 70.48%. On instantiation the model can either be trained or loaded from previous saved weight file.

cifar-100 weights cifar-10 weights

References:

[1] Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014.

[2] Shuying Liu and Weihong Deng. Very deep convolutional neural network based image classifi- cation using small training sample size. In Pattern Recognition (ACPR), 2015 3rd IAPR Asian Conference on, pages 730–734. IEEE, 2015.