Spectral sensitivity estimation with an uncalibrated diffraction grating
We introduce a practical and accurate calibration method for camera spectral sensitivity using a diffraction grating.
Oct 19, 2025
NeuraLeaf: Neural parametric leaf models with shape and deformation disentanglement
We develop NeuraLeaf, the first neural parametric model for 3D leaves for plant modeling and reconstruction.
Oct 19, 2025
Masks-to-Skeleton: Multi-view mask-based tree skeleton extraction with 3D Gaussian splatting
Jul 11, 2025
Plant Twin
The ultimate goal of this research is the complete virtualization of plants from sensing data, i.e., the generation of a plant digital twin. This research focuses particularly on technical elements related to computer vision, and reproduces not only the shape of the plant from a set of images taken of the plant, but also its branch and leaf structure, time-series changes, etc., including occluded areas. The virtualized plant model enables simulation and mapping to genes, and is a powerful tool for automating cultivation and accelerating breeding. This project is in progress under the following JST grants: 2021/04-2028/03 JST FOREST “PlantTwin: Reconstructing plants for breeding and cultivation” 2017/10-2021/09 JST PRESTO “Three-dimensional plant structure modeling and lifelog generation for growth analysis and prediction in future cultivation”
Jul 1, 2025
🎉 Two papers were accepted to ICCV 2025!
Congrats for Yang-san and Makabe-san for your great works! See you in Hawaii🌺
Jun 26, 2025
HoGS: Unified near and far object reconstruction via homogeneous gaussian splatting
We develop HoGS, combining 3DGS with homogeneous coordinates for high-quality rendering.
Jun 11, 2025
3D Reconstruction
This research aims to establish “innovative photogrammetry based on a data-driven approach and co-design of imaging systems” for the purpose of faithful 3D digitization of real-world objects using cameras, and to realize a world where anyone can easily create high-quality 3D digital content. In this project, we will develop technologies to obtain not only the rough objects’ shape (macro shape) but also the visible micro shape of the object surface (meso shape), which is essential for faithful digital reproduction and a finer shape that expresses reflectance (micro shape). In addition, this research will establish a technology for easily digitizing real-world objects in three dimensions using simple photographic equipment through the cooperative design of data-driven and imaging systems. This project is in progress under the following JSPS grants: 2023/04-2028/03 JSPS KAKENHI (S) “Photogrammetry through co-design of data-driven 3D estimation and imaging systems”
Apr 1, 2025
TreeFormer: Single-view plant skeleton estimation via tree-constrained graph generation
We develop TreeFormer, accurately estimating plant skeletal structure from a single image.
Feb 28, 2025
🎉 A paper was accepted to CVPR 2025!
Congrats for Liu-san and Huang-san for your great work!
Feb 26, 2025
🎉 A paper was accepted to WACV 2025!
Congrats for Liu-san for your great work!
Oct 28, 2024