Computer Vision＋Deep Learning
How Far Will
AI Evolve in Distinguishing Dogs from Cats?
Computer VisionMachine LearningArtificial Intelligence
What Is Computer Vision?
A dog is a dog. A cat is a cat. Humans can instantly interpret what enters their eyes and understand what it is (or what is happening) without any special training.
Computer vision is an effort to give computers visual capabilities comparable to ours. It is research that trains computers on large amounts of images and videos using deep learning＊, enabling them to correctly recognize what they “see” through a camera.
For example, it is used in many situations: finding defective products among items moving on a factory conveyor belt, detecting and reporting dangerous behavior among people making various movements on a train platform, and distinguishing normal cells from cancer cells in endoscopy. Its range of applications is expected to expand significantly in the future as well.
Tips
Deep Learning
A mechanism inspired by the structure of human nerve cells (neural networks). It enables computers to learn on their own (machine learning) and to perform analysis, reasoning, and decision-making in ways similar to humans. It is also used in ChatGPT.
Analyzing Images with Deep Learning
Deep learning, a core technology of computer vision, is a method that enables accurate recognition of the patterns and regularities in data by learning through multiple layers (“deeply”).
Take an apple, for example. Rather than defining it in words, we show the computer many images of different apples and let it learn on its own.
The dog-and-cat images visualize which parts of the original image the AI focuses on when distinguishing dogs from cats. You can see that the AI makes accurate distinctions using only about 4% to 16% of the entire image area.
Such advances in computer vision dramatically raise the level at which AI can be utilized. For example, the ChatGPT we use today is an AI tool powered by deep learning, and it supports us through language. As computer vision continues to advance, the next step will be the emergence of embodied AI—AI with a physical body. Drones that deliver packages autonomously, self-driving cars, and humanoid robots that assist with household tasks and caregiving: a future in which we coexist with such technologies is already just around the corner.
Thinking About Social Issues from an Engineering Perspective
Data science aims to make human life more comfortable and maximize well-being, and achieving that depends on continued technological progress. Advances in information engineering—such as computer vision, information and communication technologies, and computing—provide the technical foundation that drives data science forward and helps addres.
I think the strength of Faculty of Informatics at Chiba University is that it can view data science from the perspective of engineering. The appeal of our research field is that you can immediately see the results in a visible form.
In this period of social transformation, represented by the spread of AI, I believe it is an academic and research area that can realize a new society through new technologies.
I hope you’ll take on this challenge.
Profile
Prof. Kazuhiko KAWAMOTO
Professor, Graduate School of Informatics / Faculty of Informatics, Chiba University. He graduated from the Department of Information Engineering, Faculty of Engineering, Chiba University in 1997. He completed the doctoral program at the Graduate School of Science and Technology, Chiba University in 2002. After serving as an Assistant at Tokyo Institute of Technology, working at the Graduate School of Kyushu Institute of Technology, the Institute of Managementand Information Technologies at Chiba University, and then the Graduate School of Engineering and the Graduate School of Advanced Integration Science at Chiba University, he has held his current position since 2024. Ph.D. (Engineering).
View Another Research
Improving AI Accuracy and Nurturing It into a Trustworthy Partner
Assoc. Prof. Hiroshi KERA
Developing a Camera That Can Capture What Cannot Be Seen
Assoc. Prof. Hiroyuki KUBO
Eliminating Damage from Torrential Rain Through Meteorological Data Science!
Prof. Shunji KOTSUKI
Using Human Data to Support Longer, Healthier Lives
Prof. Ryoko KATAGIRI
How Far Will AI Evolve in Distinguishing Dogs from Cats?
Prof. Kazuhiko KAWAMOTO
Aiming for a New Information and Communication Society with Energy and Environmental Issues in Mind!
Prof. Hiroo SEKIYA