art_identifcation_using_CNN

Codes for identifying arts using convolutional neural networks (CNN). In this package we provide codes for preprocessing purposes such as, resizing images and relabeling data, and CNN models implemented in TensorFlow.

Data

We used art paintings provided Painter by Numbers competionKaggle competion https://www.kaggle.com/c/painter-by-numbers/data.

General usage

1. The helper functions can be found in tools/
2. A simple jupyter notebook to preprocess images and generate binary numpy files is avalaible in data/convert_images_to_numpy.ipynb
3. A simple jupyter notebook to perform classification with TensorFlow is avalaible in models/CNN_VGGNet_for_classification.ipynb

This package is still under development and more features will be added.

Authors:

Paul Hanakata, Owen Howell, Varun Ursekar