#Japan #Tokyo #CADDi #Sansan #SB Intuitions

Major Workshop on AI Development at ECCV 2026

In an exciting advancement for the field of computer vision, Caddi, Sansan, SB Intuitions, and LINE Yahoo have announced that their researchers will be organizing the international workshop titled FOUND during the European Conference on Computer Vision (ECCV) 2026 in Malmö, Sweden. Scheduled from September 8 to 12, 2026, the event aims to address the critical theme of Foundation Data for Industrial Tech Transfer.

About the FOUND Workshop

The FOUND workshop will be an integral part of ECCV 2026 and is set to take place on September 9, 2026. It will be hosted at the Malmö Arena and Malmömässan. This workshop intends to provide a platform for researchers worldwide to share insights in the increasingly important area of AI implementation in real-world scenarios. It will focus on creating and transferring foundational data, which is essential for societal applications of AI technologies.

Significance of ECCV 2026

The ECCV is organized by the European Computer Vision Association (ECVA), having been held biannually since 1990. It stands alongside other significant conferences like ICCV and CVPR, with accepted papers published by Springer. Workshops that accompany these conferences undergo a rigorous proposal review process, making the successful selection of FOUND—particularly led by Japanese corporate researchers—a noteworthy achievement.

Workshop Themes and Objectives

The current landscape of AI technology presents both groundbreaking advancements and unique challenges. Despite the significant progress made by foundation models in computer vision, practical operational issues including long-tail distribution, domain shift, and constraints related to privacy and safety have raised concerns about reliability in real-world applications. To bridge this 'last mile', there is a pronounced need for specialized datasets that reflect real-world conditions.

Key Themes Include:

• - Collection, curation, and labeling of foundational data in non-internet and domain-specific areas.
• - Self-supervised, semi-supervised, and weakly supervised learning strategies for domain-based data.
• - Development of synthetic and simulation data to fill domain gaps and cover rare corner cases.
• - Addressing domain shift, out-of-distribution (OOD) challenges, and long-tail distribution robustness in industrial implementations.
• - Data and transfer strategies for systems with physical presence and interactive capabilities, aligned with the themes of Physical AI and World Models.
• - Implementation-consistent evaluation protocols and responsible data governance for real-world AI.

Invited Speakers

The workshop will feature distinguished speakers such as Iro Armeni from Stanford University, Marc Pollefeys from ETH Zurich/Microsoft, and others leading in the fields related to AI technologies.

Building on Previous Success

The FOUND workshop follows the successful inaugural event held at ICCV 2025 in Honolulu, which fostered collaboration between industry and academia to explore domain-specific foundational data construction and problem-solving. There was notable interest from various Japanese companies, highlighting a commitment to advancing AI technologies relevant to specific industries.

Organizational Support

LIMIT.Lab, an international research community that bridges academia and industry, is the backbone for this workshop organization. Along with the core companies, additional academic partners include the National Institute of Advanced Industrial Science and Technology (AIST), the University of Tokyo, Ericsson, Qualcomm AI Research, Woven by Toyota, and Google Research, creating a robust collaborative environment.

Looking Ahead

The companies involved are committed to ongoing contributions to major international conferences, ensuring that Japan remains a leader in the computer vision field. By hosting the FOUND workshop, they aim to illuminate the pivotal role of foundational data in next-generation AI research, creating opportunities for innovative developments.

The collaborative efforts of Caddi, Sansan, SB Intuitions, and LINE Yahoo during ECCV 2026 represent a major step forward in fostering dialogue and sharing knowledge within the global research community, ultimately working towards effective solutions in real-world applications of AI and computer vision.