COMPUTER VISION: FROM PIXEL TO INFORMATION
Yiğit KULABAŞ, PhD
Redesign Business CEO & Co-founder
Computer vision has already become a part of everyday life. Computer vision is changing and will continue to change almost every sector from transportation to agriculture, from education to health, from defense to manufacturing and even space research.
I have always been interested in the wonderful relationship between the eye and the brain. We receive 11 million bits of data per second through our senses and 90% of it is collected by the eye. In a sense, the path to the brain passes through the eye.
Seeing is an ordinary ability for most us. However, it is very difficult to comprehend the complex mechanism behind. The eye has many abilities. It does not only look, but it also focuses, differentiates, prioritizes and eliminates. On the other hand, the brain makes sense of, classifies and records images.
For many years, the world of science and technology has been working to bring this natural skill of humans and animals into the computer environment. Accomplishments achieved in the last few years are remarkable.
This special field is called “Computer Vision” in English. We are talking about a subject overlapping artificial intelligence in many ways. Therefore, I prefer to call it ‘Computer Vision’ from now on.
IMAGE AND THE DIGITAL WORLD
We witnessed the great transformation of the imaging world in the nineties. Photographs and videos were quickly digitized. From the beginning of the millennium, with mobile phones being able to take photos and videos, the term pixel has become a concept that is used not only by experts but by everyone.
Today, there are many fields that combine the digital world with the image. The easiest way to define these fields and express their differences is to look at the inputs and outputs.
If our input and output are both pixels, we call this field “image processing technologies.” If our input is a definition and our output are pixels, it is possible to call this field “graphic technologies.” When our input are pixels and our output is a definition, the field is called “computer vision.”
FROM PIXEL TO INFORMATION
We can also define computer vision as “the ability to perceive what is seen.” On the journey from pixel to information, computer vision transforms the image into information and findings by analyzing a picture or a video.
Let me explain with an example. Let’s say we have a black box named computer vision. You throw a picture from one side and get a text defining the picture from the other side. There are many sentences in the text. At the end of each sentence, there is a probability rate. There are two people in this picture (100%). One is female, the other is male (100%). The female is young and in her 30s (95%). She is wearing a wedding dress and holding a flower bouquet (100%). She is wearing make-up (100%). She is happy and smiling (95%). The man next to her is young and tall (100%). He is wearing a tuxedo (100%). He is happy and smiling (95%).
They are holding hands (100%). The photograph was taken outdoors, on a sunny day (95%). They are standing on grass (95%). There are cherry blossom trees behind them (95%). There are hundreds of pink flowers on the trees (100%). This is a wedding photo (98%).
When we look at a photograph, there are dozens of details that we see but do not even talk about. Computer vision turns these into a list. And thanks to these lists, automobiles, machines, toys, streets, gadgets and guns can start to “see.”
INTERSECTION POINT
Computer vision is nurtured from many important fields, mainly informatics, physics, biology, psychology, mathematics, electronics, imaging and engineering.
In the field of biology, areas such as brain-eye relationship and neurobiology stand at the forefront; whereas from the psychology perspective, cognitive functions, perception, comprehension, trial and error, learning and decision making are the main themes. The areas related to physics are more about optics and imaging. The contribution of mathematics is in statistics, geometry and optimization.
In the field of IT and engineering, the list is much more crowded. Computer vision is linked closely to popular topics such as artificial intelligence, machine learning, augmented reality, virtual reality, sensors, internet of things, e-mobility, robotics.
There is no doubt that there is a close connection between machine learning and computer vision. Most of the basic computer vision skills such as object recognition, face recognition, face comparison, emotion detection, inappropriate content and logo recognition are realized through machine learning. With the help of functions such as pattern recognition, prediction and modeling, pixels turn into information. Databases containing tens of thousands of images contribute a lot to the learning of these systems.
FROM TODAY TO TOMORROW
Computer vision has already become a part of everyday life. We are familiar with different areas of use, from motion-sensing game consoles to mobile phone cameras, from face recognition applications to security systems. There is also an important relationship between computer vision and autonomous cars, one of the important developments of the near future. Computer vision is changing and will continue to change almost every sector from transportation to agriculture, from education to health, from defense to manufacturing and even space research.
