Tech

Human vision – a challenge for AI


Human vision - a challenge for AI

If cameras are comparable to the human eye in computer vision, the code is definitely the brain, according to Michael Felsberg. Behind the code, there’s a lot of high-level math. Credits: Thor Balkhed

Achieving diversity in human vision is one of the major challenges facing AI research. In most cases, we are better than machines at understanding the world around us. But machines are catching up — slowly but surely.

“Within a day, we humans can go from driving a car to freediving, continuing to read the newspaper and navigating through a jungle — all with little effort. For a robot, do similar things are now impossible,” said Michael Felsberg, a professor at Linköping University and one of Sweden’s leading researchers in computer vision and artificial intelligence (WHO).

We humans are able to do all of this, and more, largely thanks to our vision. Estimates say that about 80% of our impressions come to us by our looks. It is the most important sense to perceive what is happening around us. Michael Felsberg’s research mainly focuses on so-called artificial vision systems, whose aim is to make computers look as good as humans.

“Biological systems work simply. Humans have considerable cognitive and analytical skills in general, skills that we want to simulate in computers. Today, we can build engineering systems that perform a particular task well, such as self-driving car. But if in the future we want to be able to collaborate with robots, they have to be able to see and understand exactly what we see,” said Michael Felsberg.

Imitating human vision seems easy at first glance. When research on AI began, the feeling was that computer vision would be solved with a simple camera — maybe a summer vacation project. Now, nearly 60 years later, computer vision in general has evolved into one of the most prominent challenges in AI research.

Code is the brain

Michael Felsberg and his colleagues tested many of the solutions they developed in the vision laboratory at Campus Valla in Linköping. For example, between giant glass walls, self-driving drones and small self-driving cars equipped with advanced sensors and cameras are being tested. But the actual brain in computer vision is behind the lens.

“The camera is just a light sensor; it can’t do anything else. The actual work is done by the code and software behind the camera. So are humans: the eyes perceive light and the brain does the work,” says Michael Felsberg.

There have been many attempts to simulate the human brain – with mixed results. Nowadays, a machine learning method called study carefully often used. Simply put, it means that the computer learns its models organized in a neural network from a large amount of data. Algorithms are fed with huge amounts of data, analyzed at many levels. This sounds complicated, and it is. The truth is that no one can say exactly what will happen during each activation in a deep network.

Michael Felsberg draws parallels with brain:

“When you scan the brain, you can see which part of the brain is active in different stimuli. But we still don’t know what’s really going on and how thoughts form in the brain. Deep learning works in a way We see that it works, but not how it works in detail,” he said.

Human vision - a challenge for AI

Autonomous vehicles and drones are some of the current application areas for research led by Michael Felsberg and his team. Credits: Thor Balkhed

The way forward

But why is it so hard for computers to see what we see? The answer lies in the ability to quickly adapt to different situations and feedback loop between our perception of our surroundings and our capacity for constant positive perception.

Looking out through a dirty window is an everyday example of a situation where computers have a hard time but we humans manage with ease. We immediately saw what was going on outside the window, although our view was slightly obscured. Otherwise, the computer will automatically focus on the dirty part of the frame first. But once it found the right focus — on the outside scene — it still wouldn’t fully understand what was going on, because some of its vision was covered with dirt.

However, there are areas where computers have seen better than humans – especially when it comes to accurate calculations and assessments of distances, temperatures, and patterns. In these cases, computer vision can supplement our own vision, rather than draw our own conclusions and act on them.

Michael Felsberg said: “A technical system works well as long as everything is as expected. But when faced with something unexpected, it will have problems. We have to work to make the system strong. stronger”.

But software development can surpass the flexibility of human vision waste of time. And according to Michael Felsberg, research takes time if it is to be powerful. Science is a process, and each new research paper adds another small piece to the big puzzle. Breakthroughs that give research a leap forward are rare.

Michael Felsberg said: “General situational awareness in computers may persist in our lifetimes. But making the connection between cognition and general situational awareness in computers is probably a long way into the future. “.

When general computer vision exists, he believes there will be various applications, such as social robots, safer autonomous vehicles, and more efficient manufacturing. But AI is not uncontrollable. Many areas of use pose a risk of invasion of personal privacy when large volumes of personal data are processed.

For this reason, Michael Felsberg and his research team are focusing on how AI can provide additional insight into how we can prevent climate change additionally:

“Climate change is one of humanity’s greatest threats. Using advanced computer vision, we will be able to quickly analyze large swaths of land and their importance to the climate. It will take a few years for humans to map out manually that can be completed in weeks with the help of AI.”


Data sets are the bridge between human vision and machine learning


Quote: Human Vision — a Challenge to AI (2022, October 22) retrieved October 22, 2022 from https://techxplore.com/news/2022-10-human-visiona-ai .html

This document is the subject for the collection of authors. Apart from any fair dealing for personal study or research purposes, no part may be reproduced without written permission. The content provided is for informational purposes only.

news7f

News7F: Update the world's latest breaking news online of the day, breaking news, politics, society today, international mainstream news .Updated news 24/7: Entertainment, Sports...at the World everyday world. Hot news, images, video clips that are updated quickly and reliably

Related Articles

Back to top button