Projects
TL; DR: A dual-pixel dataset containing multiple types of information and an NN-based dual-pixel image simulator.
F. Li, H. Guo, H. Santo, F. Okura, and Y. Matsushita. Learning to Synthesize Photorealistic Dual-pixel Images from RGBD Frames. ICCP 2023. [code & dataset]
TL; DR: A new formulation to model the linear regression problems with unknown correspondences. A remarkably simple yet effective optimization algorithm is also presented.
F. Li, K. Fujiwara, F. Okura, and Y. Matsushita. Shuffled Linear Regression with Outliers in Both Covariates and Responses. IJCV (published in 2022, IF=19.5). [paper]
F. Li, K. Fujiwara, F. Okura, and Y. Matsushita. Generalized Shuffled Linear Regression. ICCV 2021 (oral). [paper] [supp] [code]
TL; DR: There are 24 ambiguities of the PCA-based canonical poses of point clouds! Correctly identifying them leads to much better performances in deep rotation-invariant point cloud analysis.
F. Li, K. Fujiwara, F. Okura, and Y. Matsushita. A Closer Look at Rotation-invariant Deep Point Cloud Analysis. ICCV 2021. [paper] [supp] [code]
TL; DR: Geometric and probabilistic point cloud registration algorithms are developed independently so far. We propose a framework that unifies them together for an information-geometric perspective.
F. Li, K. Fujiwara, and Y. Matsushita. Toward A Unified Framework for Point Set Registration. ICRA 2021. [paper]
TL; DR: Adaptive Bayesian filter based on Bingham distribution for pose estimation in SE(3) .
F. Li, GAG. Ricardez, J. Takamatsu, and T. Ogasawara. Adaptive Bingham Distribution Based Filter for SE(3) Estimation. ICRA 2019. [paper]
