What Is Computer Vision
What is Computer Vision, and what will its impact be on our world?
Ever wonder how computers show images of cats, when you type in 'cats'? How does it know what a cat looks like? For years, techy types around the world have been trying to make computers 'see' the content of digital images. This field of study is called computer vision, which is often abbreviated to CV.
How Does It Work?
Computer vision uses an AI neural network, much like a human brain; but unlike a human brain, computers work in binary, and can only add two numbers together. They do this millions of times in order to perform any task, including accurately recognising similar images. To undertake this task properly, the AI system needs large quantities of data, which it uses to recognise patterns.
Whilst computer vision has advanced hugely since it was pioneered in the 1960s, it is still inherently flawed. Take the example of Chihuahuas and muffins. Google (and all search engines) can't always tell the difference between the two, whereas humans can easily distinguish. Like humans, the more examples that have been seen by the computer, the more chance it has of identifying the image correctly, as the signal is understood through comparison with a set of references. Both human and computer vision systems evaluate whether an image is sufficiently similar to previously seen images to be another example of a Chihuahua; or a muffin. Humans and computers can both recognise patterns in images, but computers cannot explain the reasons for their decisions.
Because even young children can tell the difference between these two things, the problem of computer vision appears pretty simple. However, it remains unsolved because of the fairly limited understanding we have of biological vision, combined with the almost infinite variations of 'things' in our world.
Currently, methods are being developed by thousands of intelligent and creative minds around the world, all attempting to reproduce the amazing capabilities of human sight. The way that the brain processes the information it receives from the eyes, and the workings of the eye itself, must be fully understood in order to improve computer vision.
The Possibilities Of Artificial Intelligence Are Endless
Once artificial intelligence has advanced enough, the possibilities will be extraordinary, but the downside is that many jobs could be at risk. Take for example, quality checking, which could be performed entirely by computers, negating the need for staff. Robots could replace soldiers, identifying and shooting enemies.
Computer vision is already in the early stages of being used for medical diagnosis. In fact, in 2017, Stanford computer scientists reported some success with a computer vision tool that used artificial intelligence in the diagnosis of skin abnormalities; an early indicator of skin cancer. Eventually, software will be used regularly to detect cancer, along with many other diseases. The computer's accuracy rates will probably be higher than those of a doctor, but unlike a doctor, a computer cannot explain to the patient the reasons for its decision. Which brings about the question, who would you rather trust; a doctor or a computer?