Matias Turkulainen

MSc Robotics, Systems and Control

matiasturkulainen@gmail.com

MSc student in Robotics at ETH Zurich interested in Computer Vision and Computer Graphics.

ETH Zurich: Zurich, Switzerland; MSc Robotics, Systems and Control 2021-2024 Semester Project: Object 3D shape reconstruction from RGB-D images using trained shape priors. Created a custom multi-view object RGB-D dataset of textured ShapeNet models using Blender and an auto-encoder capable of single or multi-view voxel reconstruction with PyTorch. A small demo is available.

Relevant Coursework: Vision Algorithms for Mobile Robots 5.5/6, Image Analysis and Computer Vision 5.25/6, Robot Dynamics 5.25/6, Introduction to Machine Learning 5.25/6, Computational Models of Motion 5.25/6, Computer Graphics, Mixed Reality 5.75/6, Machine Perception, and Seminar in Advanced Topics in Vision 5.75/6.
University of Glasgow: Glasgow, Scotland; BEng (Hons) in Mechanical Engineering with Aeronautics 2017-2021 Grade: 1st class honours. Cumulative GPA: 20.5/22. Bachelor’s project: “Comparison of SLAM Algorithms for Drone Operations” using ROS + Gazebo simulations.

Research Intern: VTT Technical Research Centre of Finland April 2023 - Working on neural inverse rendering. Project related to fusing rgb + hyperspectral imaging into hyperspectral radiance fields for material classification in 3D. Contributing to the open-source Nerfstudio project. Also worked on camera pose refinement with BARF-nerfstudio and the various problems of registering poses of rgb and hyperspectral images.
Research Assistant: Computer Vision and Learning Group (VLG), ETH Zurich Feb 2022 - July 2022 Project related to human pose estimation from images. Data set capture of social interactions using Microsoft Hololens and data processing of OpenPose and AMASS 2D/3D keypoints with Python. The work led to a publication: EgoBody.
Drone Systems Trainee: Nokia Jun 2020 - Aug 2021 R&D intern at Nokia working with LTE connected drones. System level testing and analysis of experimental hardware/software. Worked on camera based precision landing with Jetson Nano and OpenCV.

gsplat: https://docs.gsplat.studio Together with researchers and students at Berkeley BAIR lab, we modularized and reimplemented the influential "3D Gaussian Splatting for Real-Time Radiance Field Rendering" paper from SIGGRAPH 2023. gsplat contains Python bindings for various lower level CUDA kernels allowing easier experimentation and research in conventional Python based workflows.
Nerfstudio: https://docs.nerf.studio Maintainer and contributor to the open-source NERF project which is trying to bring the latest tech in 3D neural reconstruction to devs and other users. My work includes bug fixes, help with issues and questions, and some major overhauls like this PR. I also worked on camera optimization with BARF-nerfstudio and created a template for other devs to expand on.