London-based machine learning start-up Deep Render has raised £1.6m to power the further development of its image compression technology.
The firm, which spun out of the robotics lab at Imperial College London in 2017, has created what it calls biological compression technology. It claims the algorithm is context aware; reducing file-size while keeping detail in areas of the image where the eye is likely to be drawn.
“We’re not trying to make the original software better, but replace it. Effectively, we are burning the existing compression technology to the ground; rewriting, redefining and reinventing the entire domain,” said co-founder Christian Besenbrush.
Co-founder Arsalan Zafar said human eyes have evolved to care about certain colours and properties of the natural world, a feature replicated in its algorithm.
“We are more sensitive to the colour green as it represents fertile areas with food and water; we notice the slightest of movements in still scenes, as this helps us flee from sneaking predators. Accounting for these evolutionary instincts improves visual quality but teaching a machine to do so has been incredibly complex until now.
“Our technological breakthrough represents the foundation for a new class of compression methods.
The latest round was led by Pentech with participation from Speedinvest and will be used to fund further development of the compression technology alongside new hires.
Eddie Anderson of Pentech added: “Image compression is at the heart of almost everything we do online, at home and in business, yet we’re stuck using ill-equipped and antiquated algorithms that won’t, and can’t, meet future needs.