Overview
Neural-style is a Torch-based implementation of a neural algorithm for artistic style transfer. It leverages Convolutional Neural Networks (CNNs), specifically VGG-19, to extract and combine the content of one image with the style of another. The algorithm separates style and content representations, allowing users to render content images in the style of famous artworks. The user can control style scale and blend multiple style images. It utilizes optimization algorithms like L-BFGS and ADAM. Dependencies include Torch7, loadcaffe, CUDA, and cuDNN. The tool uses Lua scripting. OpenCL backend is also supported. NIN Imagenet models can be used for smaller memory GPUs.
Common tasks