Have you ever wondered how self-driving cars see the world or how a virtual character in a video game recognizes objects?
This is all due to a technology called computer vision, which gives computers the ability to ‘see’ and understand images.
However, these systems sometimes struggle to recognize objects that are blurry or far away, just like when you squint your eyes trying to read a distant sign.
An interesting solution to this problem has been proposed by Lin Gu, a researcher at RIKEN, a scientific research institute in Japan.
Gu was originally a therapist, and his background helped him think about this issue in a unique way.
He noticed a similarity between AI’s difficulty recognizing objects at different distances and a peculiar human vision condition called Alice in Wonderland syndrome, where people perceive objects as larger or smaller than they actually are.
Just like humans perceive objects as the same size no matter how far away they are, Gu wanted to develop an AI that could do the same. He decided to use an approach inspired by how our brains form memories.
His team developed a model that randomly changes the quality of an image (for example, making it blurrier or changing its resolution), then looks for features that stay the same even after all these changes.
Gu and his team’s method is different from the usual ways to train AI. They change the quality of the image, rather than masking parts of it or changing its contrast.
This new method helps the AI to understand what objects are in the image and where they’re located without any help from humans. It’s a more efficient way of teaching AI to understand and process critical details in images.
Gu believes this new approach to computer vision will be key in the development of futuristic technologies like cybernetic avatars.
He is currently working on a project that aims to create a digital version of a government official that can interact with citizens!
In addition, this new AI approach is being used in an imaging technique called terahertz imaging, which uses electromagnetic waves to create pictures without causing any harm.
This method has exciting potential in fields like medicine, security, and material science.
Gu and his team are collaborating with researchers from Oxford University to create the next generation of these imaging devices, using AI to improve their quality and resolution.
So, Gu’s work could revolutionize how AI ‘sees’ the world, paving the way for smarter self-driving cars and more interactive digital avatars. Who knew the world of AI could be so fascinating?
