As my postings in this blog show, I got interested in image processing by neural networks through style transfer, two and half years ago. Style transfer takes two images, a content image and another image to act as a style model, and then recreates the content image using color and …
In May 2000 I participated, as an outsider to the field, a conference in Jewish Studies, with a presentation called “Slimy stones and philosophy”. It looks like the title drew quite an audience and the room was fully packed with people. This started my decade and a half as a …
New innovative GAN implementations are published every week. A particularly interesting one is CycleGAN, which promises to learn to do transformations, similar to what pix2pix does, but without needing pairs of corresponding images. In other words, it claims to be able to transform horses to zebras using a training set consisting …
There is much hype out there concerning style transfer, but there is very little serious discussion about how all this relates to art. There are claims of being able to produce art, or that art as we know it is now past, while it is not at all obvious that …
Pix2pix is a brand-new tool which is intended to allow application-independent training of any kind of image transform. All that is needed is a dataset containing image pairs, A and B, and one can train a network to transform to either direction. Possible applications include colorization, segmentation, line drawing to …
When I found neural style transfer in September 2015, I was not looking for something to instantly convert my photos into works of art, but rather new tools to use within an artistic workflow. This, I think, puts me in a minority amongst neural style transfer users and developers. Before …
A couple of years ago I participated a research project on Adam traditions in different religions from fresh points of view. For my own paper, I proceeded from a story in the Hebrew “Book of Raziel”, which tells how Adam is confused, full of grief and anxiety, when he is …
In my two previous posts, I experimented with neural-style by taking the fully connected layers into use. This resulted in something quite different, which I have provisionally called neural-mirage. Neural-mirage looks at the uppermost fc layer, the most abstract classification of what the network thinks it sees in the image, and …
In my previous post, Controlling image content with FC layers, I described how I had modified neural-style to optimize content based on the classifications obtained from the top-most, fully connected layers (which neural-style does not use). I had found that the modified program produced visually quite interesting results, but I …
A few days ago I got an idea to generate images based on fully connected layers of the VGG19 net. As the these layers should contain abstracted information on what the net sees in the input image, perhaps in this way I could make neural-style to generate images based on …
