Victor Kitov (MSU)

Neural style transfer - introduction and applications

Abstract: The talk is devoted to neural style transfer - a technique of rendering one image in the artistic style of another image using deep neural networks. General approach of extracting and adding style is given together with recent algorithms, allowing to do this in real time and preserve important information on the stylized image. This is followed by discussion, how presented technique could be applied to objects of arbitrary nature and used for better training of machine learning models with intelligent data augmentation.