Takeo Kanade (Hyōgo, Japan, 78 years old) speaks fluently about artificial vision to which he has been dedicated for more than 40 years. This scientific discipline allows you to see a soccer match from the point of view of the ball or a tennis match through a hawk’s eye. The fundamental algorithms that Kanade has developed with his colleague Bruce Lucas, called the Lucas-Kanade method, help computers and robots understand moving images. His work has also contributed to improving robotic surgery, autonomous driving and facial recognition. “In the future, robots will be better than humans, in one way or another,” he admits. Thanks to his mathematical developments, Kanade has received the Frontiers of Knowledge Award from the BBVA Foundation, in the Information and Communication Technologies category.
The researcher received his doctorate in Electrical Engineering from Kyoto University in 1974. He discovered his passion for engineering when he went fishing at the age of 5 and made his own hook. He is a professor of Computer Science and Robotics at Carnegie Mellon University in Pittsburgh (USA) and founder of the Technology Center for Quality of Life, which he directed between 2006 and 2012.
Ask. Do you think robotic vision will one day match human vision?
Answer. Yes, and at some point it may be better than that of humans. In fact, in some areas they are already better, such as in the case of computer facial recognition. For a long time it was thought to be an area of great advantage, as we are good at recognizing people we know. However, if we meet someone new in an unfamiliar environment, we often lose sight of them.
P. Which of the two is least likely to fail?
R. The human being. Self-driving cars can see up to 200 meters in all directions and recognize the location of other cars, pedestrians or bicycles very precisely, but we have a better understanding. We can have some expectations about what is happening to the car in front of you. You can also recognize if you are driving near a school, and anticipate children who may cross the road. Computers are trying to get to that level of understanding, but for now it’s not that good; and it is essential to avoid accidents.
P. Will driving be 100% autonomous one day?
R. Within ten years or even less. However, people have to be convinced. It is like the use of the car, which although it causes accidents and deaths, produces such a great benefit that, as a society, we accept it.
P. In the world of artificial vision, what are the challenges of the future?
R. Can you be in a world that is mapped from the real world? That’s the next level. I call it virtualized reality. Virtual reality is now not virtual: it is a real world that is virtualized. There is also the possibility of interacting with the environment and with the people in the virtual world. That someone can see themselves reflected in a mirror within the virtual world. Another challenge is teleportation, but without decomposing your body and sending it to the other place, like in the movie Star Trek. It is done with tools, like drones, that give you a reaction force in your legs through your body so that you can teleport, visually and acoustically, and that you can physically interact in real time.
P. How can you prevent malicious use of deepfakes (fake videos)?
R. I feel partially responsible for it. In 2010, I made a video of President Obama speaking in Japanese with images generated from my face. I thought it was a prank video. You can’t fight the deepfakesthe only thing that prevents it is our integrity.
P. But if technology can turn against us…
R. It is used for the purpose of deception, but the technology cannot know what its objective is. A type of technology can be created and used for purposes other than those for which it was developed. You have to be smarter, faster than her and be informed. The watermark, for example, when you use it and it is known, at that very moment an instant way to delete it has already been created.
P. How can computer vision be used to improve the quality of life of people with disabilities?
R. It is about developing quality of life technology, as we do in our center. Its essence is to increase the independence of people with disabilities or older people. It is not about robots doing everything, but the opposite. My formula for the perfect robot equals what you want to do minus what you can do. That is, compensate for the part that the human cannot do so that he can do it together with the robot. Furthermore, in cases of rehabilitation or education, the perfect robot must do a little less so that the person regains motivation and ability.
P. How have your works contributed to surgical precision?
R. Robot surgeons can use more developed sensors than human surgeons. Human sensors are very limited, we do not have acoustic sensors or multimodal sensors. Before a surgical operation, androids can, for example, detect the location of a tumor, its shape or size, with X-rays or an MRI, and at the time of surgery, they use image sensors.
P. What has origami contributed to your career?
R. The essence of my theory, A Theory of Origami World (Origami World Theory, in English) is that the perception of the three-dimensional shape of an image must derive from a mathematical explanation and not from the result of learning. If, for example, you draw a box, there are five more possible shapes that are generated with exactly the same image, but are different from the box. When I give a talk, for example, I joke with the audience that the conference room and the building could be different shapes and the audience looks around and imagines it.