Clever computers can ‘see’ with radar and tell objects apart in the environment around it
Despite computers performing increasingly impressive feats, when it comes to recognising real world objects, they have fallen short.
But the machines are learning that just like the robots of sci-fi blockbusters, being able to tell one thing from another is a key skill.
A team of UK researchers has created software which has taught itself to recognise objects around it, with impressive accuracy.
Called the RadarCat, short for radar categorisation for input and interaction, the software scans the world around it uses radar to scan the world around it.
Tests in the computer lab have shown the platform, created by researchers at St Andrews University in Scotland, can correctly identify everything from a sponge to a block of wood and even tell if a glass has water in it or not.
By using a machine learning approach, the software learns from a few examples in order to work out the parameters.
The approach is reaching increasingly impressive milestones in the real world, such as Google-backed machine learning platform Deep Mind being able to navigate the London Underground network, a task which regularly stumps tourists.
Objects are placed in front of a sensor which detects changes to a radar field and works out what it is ‘looking’ at.
It can even identify different body parts.