Yusuke Mukuta

About

Yusuke Mukuta is a researcher advancing the frontiers of 3D computer vision and generative modeling. His primary research areas include 3D point cloud analysis, instance segmentation, and long-term video prediction. Mukuta’s major contribution to 3D instance segmentation, detailed in his 2019 work "Rethinking Task and Metrics of Instance Segmentation on 3D Point Clouds," critically re-evaluated the prevailing practice of splitting point clouds into small regions for processing. By identifying limitations in existing models and metrics, he laid groundwork for more robust, scalable approaches essential for autonomous cars and robots. This work has garnered 5 citations, reflecting its foundational role in the field. In parallel, Mukuta addressed the challenge of predicting multiple plausible futures from video input. His 2019 study "Long-Term Video Generation of Multiple Futures Using Human Poses" innovatively leveraged human pose information to generate diverse, long-term future sequences—a significant step beyond single-future, short-horizon predictions. This work, with 2 citations, holds promise for applications in autonomous driving and robotics where anticipating varied behaviors is critical. Mukuta’s research demonstrates a commitment to solving fundamental problems in perception and prediction, making him a notable figure in these rapidly evolving domains.

Research Focus

Key Achievements

Top Papers

1. 1
   Rethinking Task and Metrics of Instance Segmentation on 3D Point Clouds

   5 citations · 2019
2. 2
   Long-Term Video Generation of Multiple Futures Using Human Poses.

   2 citations · 2019

Key Collaborators

Contact & Links