Tianshu Kuai

Previously, I graduated from the Engineering Science program at the University of Toronto, majoring in Robotics. I was fortunate to work with Prof. Steven L. Waslander at TRAILab on 3D LiDAR object detection.

I am interested in the fields of 3D vision, and my current research focuses on 3D scene understanding and manipulation. I also work on real-world image restoration and enhancement. Previously, I worked on self-supervised representation learning for 3D perception in autonomous driving. Aside from vision, I am also interested in robotics research.

I am actively looking for a PhD position for fall 2024.

Email  /  CV  /  Google Scholar  /  Github  /  LinkedIn  /  Twitter

A novel pipeline for building 3D animatable models for articulated and deformable objects from monocular videos without any shape prior or template.

A self-supervised 2D-to-3D representation learning framework for LiDAR semantic segmentation in autonomous driving.

An end-to-end two stage LiDAR 3D object detection architecture designed to account for point cloud density variations.

Research on diffusion model based real-world image restoration and enhancement.

Research on 3D scene representation and manipulation.

Research on supervised/self-supervised LiDAR 3D perception models for autonomous vehicles.

Worked on the real-time 3D Perception system for the University of Toronto's autonomous vehicle team winning the first place in the SAE Autodrive Challenge.

Research on deep learning based models for audio signal processing and state-of-the-art methods for neural network compression, and contributed to NPU software compiler pipeline development.