This research aims to construct a technical system for “occlusion-aware 3D reconstruction,” which reconstructs 3D shapes, including occluded (hidden) areas, from images taken from a single or multiple viewpoints. The reconstruction of the occluded area is essentially an unavoidable problem in the recognition and understanding of many scenes in this world. The method we construct in this research will be a new 3D reconstruction framework that estimates the 3D shape in such a way as to complete the occluded area on the image and in space, based on prior knowledge of the structure and similarity of the target object.

Grant

• 2021/4 – 2025/3 JSPS KAKENHI (B)
  PI: Fumio Okura, Grant: 12,300K JPY

Publications

• Xinpeng Liu, Hiroaki Santo, Yosuke Toda, Fumio Okura: TreeFormer: Single-view plant skeleton estimation via tree-constrained graph generation. Proc. IEEE/CVF Winter Conference on Applications of Computer Vision (WACV’25), Feb 2025. [arXiv] [code]
• Takuto Narumoto, Hiroaki Santo, Fumio Okura. Synthesizing time-varying BRDFs via latent space.
  Proceedings of European Conference on Computer Vision (ECCV), 2024. [Paper]
• Susumu Kikkawa, Fumio Okura, Daigo Muramatsu, Yasushi Yagi, Hideo Saito. Accuracy evaluation and prediction of single-image camera calibration. IEEE Access, Vol. 11, pp. 19312-19323, Feb 2023. [Paper]
• Satoru Tsugawa, Kaname Teratsuji, Fumio Okura, Koji Noshita, Masaki Tateno, Jingyao Zhang, Taku Demura: Exploring the mechanical and morphological rationality of tree branch structure based on 3D point cloud analysis and the finite element method. Scientific Reports, Vol. 12, Article No 4054, Mar 2022. [paper]
• Fumio Okura: 3D modeling and reconstruction of plants and trees: A cross-cutting review across computer graphics, vision, and plant phenotyping. Breeding Science, Vol. 72, Issue 1, pp. 31-47, Feb 2022. [paper]