Conference Proceedings

(2024.2.19)

Keynote Lectures

Abstract: Our visual world is full of complex patterns with multiple constituents, and thus visual understanding often requires the ability to grasp the relations and structures of visual elements, i.e., which visual elements are related to each other and how visual elements are structured as a whole. Despite the remarkable progress, modern neural networks for computer vision are still limited in the relational and structural awareness of visual patterns. In this talk, I will present a series of my recent and ongoing work on high-order feature transforms that address the aforementioned limitations and enable minimally-supervised recognition and structural perception of images, videos, and 3D data, focusing on few-shot recognition, motion-aware action recognition, and object assembly.

Biography: Minsu Cho is an Associate Professor at POSTECH, South Korea, leading POSTECH Computer Vision Lab. Before joining POSTECH in the fall of 2016, he worked as a postdoc and a starting researcher at Inria WILLOW team and École Normale Supérieure, Paris, France. He completed his Ph.D. in 2012 at Seoul National University, Korea. His research lies in the areas of computer vision and machine learning, tackling problems such as visual semantic correspondence, symmetry analysis, object discovery, video action recognition, and minimally-supervised learning. He is interested in the relationship between correspondence, symmetry, and supervision in visual learning. He is an editorial board member of the International Journal of Computer Vision (IJCV) and IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) and has been serving as an area chair in conferences including CVPR, ECCV, ICCV, and NeurIPS. In 2020, he was inducted into the Young Korean Academy of Science and Technology (Y-KAST). This year, he serves as a Program Chair for KCCV 2024, Busan, Korea, and ACCV 2024, Hanoi, Vietnam.

Abstract: In the context of robots that learn from external information, acquiring visual experiences is a crucial challenge. Recent research has made it increasingly feasible to generate virtual visual data. However, there are several gaps to consider in turning this into 'experience' for behavioral learning, leading to various research approaches. This overview explores familiar techniques for data augmentation and environmental randomization known to image engineers and researchers. Additionally, it delves into methodologies that leverage experiences not easily obtained in reality. The presentation includes examples of applying these approaches to manipulation and cross-modality acquisition. 

Biography: Yukiyasu Domae earned his B.E., M.S. in Information Science, and Ph.D. in Information Science from Hokkaido University in 2004, 2006, and 2012, respectively. For a decade starting in 2008, he conducted research in Machine Vision at Mitsubishi Electric Corporation, contributing significantly to the practical implementation of 3D sensors for robots and the automation of object manipulation by Industrial Robots. His contributions led to prestigious awards such as the R&D 100 Awards in the United States and the Robot Award by Japan's Ministry of Economy, Trade, and Industry. Since 2018, he has served as the Team Leader of a research team focusing on the industrial application of robotics and AI technologies at AIST, one of Japan's largest research institutions. His achievements encompass over 40 awards for robot research and development, publication of more than 60 peer-reviewed papers and proceedings, including ICRA, IROS, CVPR, and the acquisition of over 40 patents in Japan, the United States, and other countries. Additionally, he currently holds positions as an Invited Professor at Osaka University and NAIST.

Conference Program

Technical Tour

Participants will visit the Industrial Cyber Physical System Research Building, which serves as the research and development hub for AI applications at AIST. During the visit, participants will explore environments resembling factories and convenience stores, with a focus on the industrial applications of AI, robotics, and vision technology.

The technologies showcased will include robot picking, error recovery systems through remote operation, 3D measurement systems for objects, and the estimation of force information using cameras.

Location: AIST Tokyo Waterfront (Access)

Date: February 21th, 2024    13:00 - 14:30   (Tentative)

Tentative Itinerary:

• 13:00    Meeting place and time: AIST Tokyo Waterfront Annex Front Desk

• 13:00-13:30    Introduction of AIST Tokyo Waterfront Collaboration Center

• 13:30-14:00    Tour 1

• 13:30-14:30    Tour 2

• 14:30    Dismissal

photos by D-s-yama