Harvey Mudd College FA20 CS121 Section 2: Deep Learning Classification of Art Pieces
Painting Classifier is a web app created for the purpose of classifying images. Our target audience is art students who are learning how to identify art on their own, or aspiring artists who are curious about what their painting would be classified as. The web app uses trained convolutional neural network models to classify the Style, Artist, and Genre of an art piece.
Once the models have classified the painting, the app displays the results and the confidence level for each result. In addition, it uses a Google Custom Search API to find other similar paintings in the same categories.
Click here for detailed instructions on how to use the site.
Painting Classifier is currently hosted on the Google Cloud Platform.
All of our models are hosted at the FA20-CS121-fastai repository. You can see our data preparation and training scripts.
We owe a huge thanks to our colleagues in CS121, Professor Z Sweedyk, and grutor Macy Mills for being immense resources during the development process.
We also thank the authors of the ICIP2016 paper "Ceci n’est pas une pipe: A Deep Convolutional Network for Fine-art Paintings Classification" for providing the inspiration for this project. In particular, our gratitude goes to C.S. Chan for providing the formatted source data for our training, which can be accessed at the cs-chan/ArtGAN repository.
Finally, we thank the WikiArt team for compiling such a rich database of visual artworks that can be enjoyed by the public and used by researchers alike.