November 11, 2020

University of Helsinki sharpens vision of smartphones

An illustration of a smartphone against a blue background.
The benefits of highly accurate colour detection are myriad.
spexel.ai

Researchers at the University of Helsinki are developing an artificial intelligence-based solution that turns ordinary smartphones into cameras capable of detecting shades of colours extremely accurately.

The solution combines an ordinary smartphone, an ancillary attachable device and cloud-based computer vision algorithms to enable the smartphone to determine the colours of photographs to an extremely high degree of precision.

Such information can benefit consumers and professionals alike by allowing, for example, interior designers to determine the exact shade of a particular colour or online shoppers to find a shirt that matches the exact colour of their favourite jeans.

“The phone peripheral is cheap and compatible with basically all smartphones,” said Arto Klami, assistant professor at the University of Helsinki. “Consumers don’t need to buy a separate device.”

One of the two doctoral students of computer science who developed the algorithms, Chang Rajani, pointed out that while there are other devices capable of detecting colours at comparable accuracy, they need to be pointed very close to the object examined.

“Our solution makes it possible to determine the colour from further away,” he said.

Previously unseen details

Accurate colour detection, however, is only part one of the product development project, called spexel.ai. The solution also has the potential to bring hyperspectral imaging within reach of consumers, as the algorithms are capable of converting images taken with the ancillary device into hyperspectral images to reveal details otherwise indiscernible to the naked eye.

“Our vision is to bring hyperspectral imaging to consumers,” added Ville Kurri, the commercial lead of the project.

The vision could become reality due to a key difference in the development of hyperspectral images: while current techniques create the images independently from start to finish, the solution forwarded by the research team produces them from regular photographs taken with the ancillary device with the help of the cloud-based algorithms.

Hyperspectral imaging is today used for applications such as crop yield estimations, geographical remote sensing, and counterfeit art and pharmaceutical identification.

The developers are currently exploring the needs of potential customers while continuing with the proof of concepts and negotiations with investors. Commercialisation is similarly underway with a view to making the device and application available to consumers by 2021.

Looking for more good news? Subscribe to our newsletter

Share