from student of Institut Teknologi Sepuluh Nopember
"Van Gogh painting style on Surabaya landscapes? Kanagawa's Great Wave off on ITS building? you come to the right cave, VGG16 on free google colab GPU."
Hello, I'm tian a math student from ITS. Will be graduating later at march 2020. I have a lot interest in the NLP research field, yet solving problems in the industry with deep learning is what I mostly do, I'm working remotely as a machine learning engineer at a startup in Jakarta to fulfil my thirst of contributing to society, I'm a lifelong pattern hunter and I see deep learning as a bridge to humankind to acquire knowledge, by representing it to be "computer like data" makes it way faster to achieve knowledge beyond human mind. My goal is to bridge human understanding and computer knowledge.
Exploring pretrained weights of VGG19, VGG16, NIN models to detect the "edges" to copy the pattern / texture of a painting then apply the pattern/texture to a whole new picture. This is known as Style Reconstruction. Some of the models are rely on a good GPU, while to optimize resource, some models are CPU friendly. Which produces the best result?
This talk is using pytorch, based on a tensorflow and lua implementation of Justin Johnson's [neural style](https://github.com/jcjohnson/neural-style), which recreated from a [paper](https://arxiv.org/pdf/1508.06576.pdf) from Leon A. Gatys, Alexander S Ecker.