Michal Parusinski

Improving HiDPI support by upscaling icon themes with AI – Part 1

I am starting a new open source project. The goal is to solve a improve HiDPI support by upscaling icon themes with AI: old software usually rely on icon themes in raster format. These are not suitable for HiDPI screens. By relying on formats like JPG, PNG, …, low resolutions icon themes (e.g. 32×32) do not render well on HiDPI screens. The naive solution is to simply is to do pixel doubling, but this results in an ugly pixelated look. My new project will use some tools from artificial intelligence to do better.

The way to solve this issue using artificial intelligence is to use super resolution deep learning models. These neural networks rely on images subtle statistical relations between the pixels of an image. For instance corner pixels of an object have similar colour. Another example is that icons frequently feature lines and curves.

A series of blog posts will detail the progress of the project. You can start reading about the progress over here specifically on the topic of gathering data.

, , ,
Problem solving, innovative and creative with a fondness for challenging myself, Michal Parusinski have 20 years of experience in the information technology industry as a mathematician and a software developer Graduated with first class honours at from Imperial College London in Mathematics and Computer Science. Published two scientific papers relating to applied machine learning and submitted one patent.

One response to “Improving HiDPI support by upscaling icon themes with AI – Part 1”

  1. Improving HiDPI support by upscaling icon themes with AI – Part 2 – Michal Parusinski

    […] article is part of a series about upscaling icon themes with AI. While this page is part 2, part 1 is available here and part 3 is available […]

Leave a Reply