I worked on this project with the Statistical Modeling and Deep Learning Lab in Undergraduate Lab at Berkeley.
Using PyTorch, we deployed a VGG convolutional neural network to identify objects in images. We then synthesized a texture from a Van Gogh artwork while preserving the content of our base image.
After running style_transfer.py on our images, we can see our minimization of loss with each iteration of of our model. Running numerous iterations of our model produced the content below.