Two papers were accepted at CVPR 2026 (D2 Shibata, M2 Kohyama)

Two papers listed below have been accepted for CVPR 2026.

1. "Geometric-Photometric Event-based 3D Gaussian Ray Tracing" was accepted as a Spotlight paper at CVPR 2026.

This research proposes GPERT, a 3D scene reconstruction method leveraging the high temporal resolution of event cameras. The rendering is divided into event-based geometric (depth) estimation and snapshot-based radiance (luminance) estimation, which are integrated through ray tracing, achieving fast and accurate 3D reconstruction without requiring pretrained models or COLMAP initialization.

Authors: Kai Kohyama, Yoshimitsu Aoki, Guillermo Gallego, Shintaro Shiba

– Project: https://e3ai.github.io/gpert/
– Paper: https://arxiv.org/abs/2512.18640
– Code: https://github.com/e3ai/gpert

—

2. "AssistMimic: Learning to Assist — Physics-Grounded Human-Human Control via Multi-Agent Reinforcement Learning" was accepted at CVPR 2026.

This research formulates the imitation of close physical assistance between humans as a multi-agent reinforcement learning problem. By simultaneously learning partner-aware policies for helper and helpee in a physics simulation, the control of physically plausible and socially meaningful assistive interactions was achieved.

Authors: Yuto Shibata, Kashu Yamazaki, Lalit Jayanti, Yoshimitsu Aoki, Mariko Isogawa, Katerina Fragkiadaki (Carnegie Mellon University, Keio University, Keio AI Research Center)

*This research is a collaboration with Carnegie Mellon University, Aikogawa Laboratory at Keio University, and Keio AIC.

– Project: Coming soon