Final Project for CS231C (Computer Vision and Image Analysis of Art)
Recently, computer vision techniques have been applied to image analysis of art to better understand artistic choices, context of the art, etc. In this project, I explore applications of art forgery to woodblock prints – an art form quite unique to East Asia. Specifically, I use a CNN to detect original and replica prints for Hokusai’s Gaifu Kaisei (ie Red Fuji). I found that the model is able to distinguish originals from replicas with a relatively high accuracy of 92%.
Moreover, I attempt to visualize and better understand rules and features of prints used by the model in its classification (real/fake) of these prints.
See final report for complete write up, explanations, and analysis.