Grand Award

3rd Place

Craquelure Classification of European Regional Paintings Using Convolutional Neural Network

Robotics and Computer Science
Mason Fosdick

Sandra Crusa

"Abstract: Craquelure is the distinct patterns of cracks that form in painting. These cracks are determined by many factors, the most prevalent of these factors are the drying process, materials of the paint, and material to which the painting is applied. These factors are based on region and period. There are 4 distinct craquelure patterns from 4 European regions. These regions are Dutch, French, Flemish, and Italian. Convolutional Neural Networks (CNNs) are machine learning programs that are most notable applied to images. This is because of their unique convolutional layers. They have special techniques for processing images which make them optimal for image classification. This project applied a developed algorithm CNN to craquelure. A classification program that could tell the painting’s location of origin was desired. This was achieved, with a program that was at least 75% accurate and at times 90% accurate for determining the location of where a painting originated."

Project presentation

View Project Presentation file

Lab journal excerpts

View Lab Journal file

Research paper

View Research Paper file

One thought on “Craquelure Classification of European Regional Paintings Using Convolutional Neural Network

  1. This project is so cool. It is really neat that a computer can determine and evaluate cracks in a painting. This is such an amazing idea and application of computer data science. I really like how you explained your procedure using a flow chart. It was a complicated one too. This is a very cool way to solve data visualization problems. Just think of how many other amazing applications there are for this! Way to go!

Comments are closed.