🎉 All 3 papers selected for Spotlight presentations!
At the 32nd Symposium on Sensing via Image Information (SSII2026), held on June 10–12, 2026 at Pacifico Yokohama, Annex Hall, the Aoki Media Sensing Lab will present three papers. All three were selected from the submitted papers for the Spotlight Session (oral presentation), and will also be presented in the Interactive Session (poster). We look forward to discussing them with you on site.
Dates: June 10–12, 2026 / Venue: Pacifico Yokohama, Annex Hall (https://ssii.jp/)
1. Geometric-Photometric Event-based 3D Gaussian Ray Tracing (GPERT)
- Presentation No.: Spotlight [SO1-04] / Interactive [IS1-04]
- Authors: Kai Kohyama1, Yoshimitsu Aoki1, Shintaro Shiba1,2
- Affiliations: 1. Keio University, 2. The University of Tokyo
- Date & Time: Spotlight — Wed, June 10, 14:00–14:45 (Main Hall) / Interactive — Wed, June 10, 14:45–16:15 (IS Poster Hall)
This is the domestic (Japanese-venue) presentation of GPERT, accepted as a Highlight at CVPR 2026. Leveraging the high temporal resolution of event cameras, the method introduces an event-based 3D Gaussian Splatting framework that uses ray tracing to render the per-event geometry (depth) and the snapshot photometry (brightness) separately. Without relying on pretrained models or COLMAP initialization, it achieves strong reconstruction performance on real data. A collaboration among Keio University, TU Berlin, and the University of Tokyo.
Project: https://e3ai.github.io/gpert/ / arXiv: https://arxiv.org/abs/2512.18640
2. Overlap-aware ViT Harmonizer for Automatic Color Harmonization of Optical Satellite Images
- Presentation No.: Spotlight [SO2-03] / Interactive [IS2-03]
- Authors: Sota Sato1, Yuto Shibata1, Kazuki Ozeki1, Ryo Nakamura2, Shintaro Sho2, Yuhei Aihara2, Yoshimitsu Aoki1
- Affiliations: 1. Keio University, 2. Tenchijin, Inc.
- Date & Time: Spotlight — Thu, June 11, 15:30–16:15 (Main Hall) / Interactive — Thu, June 11, 16:15–17:45 (IS Poster Hall)
The work addresses the color (tone) mismatch that arises when combining multiple optical satellite images. The proposed ViT-based harmonizer focuses on the overlapping regions between images to automatically harmonize the color tones of satellite imagery captured under different conditions. A collaboration between Keio University and Tenchijin, a company specializing in satellite data analysis.
3. Product Assembly-State Recognition Using Latent Representations of 3D Reconstruction Models
- Presentation No.: Spotlight [SO2-07] / Interactive [IS2-07]
- Authors: Kohei Torimi1,2, Jyun-Ting Song3, Kris Kitani3, Yoshimitsu Aoki1, Takuma Yagi2
- Affiliations: 1. Keio University, 2. National Institute of Advanced Industrial Science and Technology (AIST), 3. Carnegie Mellon University
- Date & Time: Spotlight — Thu, June 11, 15:30–16:15 (Main Hall) / Interactive — Thu, June 11, 16:15–17:45 (IS Poster Hall)
The work proposes a method that uses latent representations obtained from a 3D reconstruction model to recognize whether a product has been correctly assembled (its assembly state). A collaboration among Keio University, AIST, and Carnegie Mellon University.
For full details, please see the SSII2026 program. We hope to see you at the venue.
