I am a visual artist, based in Finland, with a background in both information technology and humanities research. I have been working with neural networks to process and generate images since September 2015 when the original neural-style was released. Although most of my time is spent with neural networks based methods, I prefer to work with multiple media. I am also doing photography and video, even some painting, and most of my neural work is based on photographs, a kind of an extension of photography. To start from the photos and continue somewhere. My ultimate goal might be to simplify, to reduce images to the bare essentials, even if this goal might not be always visible in my work.
One of my favourite methods is to use one neural network to reduce photos to contours, for instance, and then to use this reduced image as a virtual printing plate, in another neural network, to produce a virtual art print which I then can print on fine art paper. Depending on the training of the two networks, often involving a significant dose of overfitting, a range of different looks can be obtained.
In “Must be one of these” below, the virtual printing plate has been composited of several pieces.
In another method, I mask a part of the images and use a neural network to complete the image, again relying on overfitting.
In another series, I have trained a model to apply a varying amount of virtual scratches to photos.
Finally, even though GANs are by far dominating the neural art at the moment, I have never totally abandoned the style transfer methods, but instead tried to extend and bend them to new directions, beyond mere style transfer. Always balancing between the transformative and the generative, it seems.