Computer Vision (Nishino Lab)

In our laboratory, we study computer vision, the science aimed at establishing the theoretical foundation and practical implementation for granting computers the ability to see, whose findings also inform human vision studies. Armed with machine learning and optics, we aim to elevate computer vision to an intelligent perceptual modality for computers, instead of merely a means for efficient consumption of images and videos by humans.

• Laboratory website

Academic Staff

Ko Nishino

Professor (Graduate School of Informatics)

Research Interests

Visual information processing / Computer vision / Machine learning

Contacts

Research Bldg #9, South wing, Room S-303
TEL: +81-75-753-4891
https://vision.ist.i.kyoto-u.ac.jp/

Ken Sakurada

Associate Professor (Graduate School of Informatics)

Research Interests

Computer Vision, Robotics, Autonomous Driving (Perception), Remote Sensing

Contact

Room I-302, 3rd Floor, South Wing, Research Building No.9
TEL: +81-75-753-5883
Spatial Context Sensing Research Project
Personal page

Ryo Kawahara

Lecturer (Graduate School of Informatics)

Research Interests

Computer Vision, Computational Photography, Physics-based Vision

Contact

Room I-303, 3rd Floor, South Wing, Research Building No.9
TEL: +81-75-753-3327

Introduction to R&D Topics

Perceiving People

The looks and actions of people embody their inner states including their thoughts and feelings, not merely their visual attributes such as gender and height. We, for instance, can easily identify the mood and intention of a person by just looking at them. In our lab, we have introduced how we can identify what are person is seeing and how a person would walk in a crowded scene. We plan to continue our work on understanding a person’s attention, intention, actions, and interactions from sight, towards realizing a rich symbiosis of artificial agents with computer vision and human beings.

Perceiving Things

We gauge a large amount of information by just looking at things and scenes around us. Instead of just recognizing the categories of objects in front of us, such as a road and a parked car, we immediately “see” information essential to interact with them such as the wetness of the road after rain and the difference in hardness between the metal body and plastic bumper of the car. Our lab has conducted seminal research on estimating rich object information from images, such as illumination, reflectance, geometry, and material. We plan to continue research on extracting such physical and semantic information from object and scene appearance.

Seeing Better

We use our two eyes to see our world in the visual spectrum, but computers need not be restricted to the same sensor limitations. In our lab, we conduct research on developing novel imaging systems that use computation as an integral part, referred to as computational photography, with the goal of capturing richer visual information to realize computational perception. Our recent work has focused on infrared imaging and subsurface light scattering modeling, leading to novel imaging systems for underwater real-time 3D sensing and transient imaging of subsurface scattering.

 

4. Understanding Spatial Context

For autonomous vehicles and robots to operate safely and autonomously in the real world, it is essential to sense the surrounding environment using sensors such as cameras and LiDAR, and to interpret the spatial context of the scene from the acquired data. Our lab focuses on developing robust spatial understanding systems that meet practical requirements, including real-time performance and privacy preservation, and that remain reliable even under limited sensing and computational resources. In particular, we study methods that leverage sensor data and temporal scene dynamics to enable accurate and efficient processing of camera and LiDAR information. For more details, please refer to this page