Genki Kinoshita

I'm a first-year Ph.D. student at the Computer Vision Lab in Kyoto University, advised by Prof. Ko Nishino. Through the research and software engineer internships at Sony Semiconductor Solutions, LY Corp.(formerly Yahoo Japan), Nintendo, and HACARUS, I have acquired a diverse and deep skill set, ranging from computer vision and machine learning to web application development.

Email  /  CV  /  LinkedIn  /  GitHub

I'm actively looking for research opportunities as an internship. If your team has an opening, please feel free to contact me!

profile photo

Research

I'm interested in computer vision and machine learning for 3D vision, in particular, scene reconstruction, self-supervised/unsupervised learning, and 3D human modeling.

Action Motifs: Self-Supervised Hierarchical Representation of Human Body Movements
Genki Kinoshita, Shu Nakamura, Ryo Kawahara, Shohei Nobuhara, Yasutomo Kawanishi, Ko Nishino
CVPR 2026 (highlight)
project page / arXiv / code

Learning human movement representations with semantically meaningful units that emerge as patterns through a bottom-up hierarchical representation learning.

REACH: Hand Pose Estimation from Room Corners
Shu Nakamura, Ryo Kawahara, Genki Kinoshita, Yasutomo Kawanishi, Shohei Nobuhara, Ko Nishino
FG 2026 (oral)
project page / arXiv / code

3D hand pose estimation from extremely low-resolution and frequently occluded views.

Camera Height Doesn't Change: Unsupervised Training for Metric Monocular Road-Scene Depth Estimation
Genki Kinoshita, Ko Nishino
ECCV 2024
project page / arXiv / code

Fully leveraging the characteristics of driving videos, i.e., car ubiquitousness and camera height invariance, enables any monocular depth estimator to learn metric scale in an unsupervised manner.

Misc.

Teaching: Exercise of Computer Programming in Electrical and Electronic Engineering at Kyoto University in 2023.