cv & curriculum

General Information

Full Name Yifei Liu
Date of Birth 23th December 1997
Languages English, Chinese

Education

  • 2020-2023
    Master's degree
    University of Zurich, Zurich, Switzerland
    • Major, Data Science
    • Minor, Informatics
    • GPA, 5.89 / 6.0, top 1%
    • Special (Guest) Student at ETH, GPA 5.83 / 6.0
  • 2016-2020
    Bachelor's degree
    University of Science and Technology of China
    • Mathematics and Applied Mathematics
    • GPA, 3.51 / 4.3

Courses

  • Computer Vision
    • 3D Vision, Grade 6.0/6.0, at ETH from CVG
    • Vision Algorithms for Mobile Robotics, Grade 6.0/6.0, at UZH from RPG
    • Deep Learning for Autonomous Driving, Grade 5.75/6.0, at ETH from CVL
    • Deep Learning, Grade 6.0/6.0, at UZH from AIML
    • Probabilistic Artificial Intelligence, 6.0/6.0, ETH from Prof. Andreas Krause
  • Natural Language Processing
  • Robotics - Optimisation - Planning and Control
  • Others
    • Introduction to Reinforcement Learning (5.75), Computer Graphics (5.5),
      Randomized and Online Algorithms (6.0), Foundations of Computing I (6.0),
      Informatics II (6.0), etc.

Projects

  • Exploring Sparse Computation with Vision Transformers, 09/2022 - 03/2023
    • Master's thesis at RPG with Mathias Gehrig, Nico Messikommer and Prof. Davide Scaramuzza
    • Grade 6.0/6.0
    • Mission - Exploring using the sparsity in the input images to process only the silient tokens in ViTs for improved inference efficiency without apparent loss in model accuracy, extending existing methods for classification to object detection and instance segmentation.
  • Building an Offline Python SLAM using COLMAP, 03/2022 -06/2022
    • Semester's Course Project at CVG supervised by Paul-Edouard Sarlin, Prof. Marc Pollefeys
    • Grade 6.0/6.0
    • Mission - extend the pycolmap Python bindings to have more control over the reconstruction process, and build a Python SLAM system that is simple, fast, and robust.
  • Efficient Spatio-Temporal Processing of Event Data, 09/2021 - 02/2022
    • Master project at RPG with Mathias Gehrig, Nico Messikommer and Prof. Davide Scaramuzza
    • Grade 5.75/6.0
    • Mission - investigate the pros and cons of two potential methods for processing event camera data, voxel-based methods such as 2D or 3D CNNs, and point-based methods such as PointNet. Combine the the point-voxel methods to improve the performance on classification and optical flow regression.
  • Multi-task Learning and Detection for Autonomous Driving, 04/2021 - 07/2021
    • An competition in ETH course Deep Learning for Autonomous Driving
    • Win the 2nd place for Lidar points 3D Object Detection
    • Mission - Adopted DeepLabv3+ model and used multi-task learning (MTL) models to achieve semantic segmentation and monocular depth estimation jointly. Tried three architectures to achieve MTL, Joint architecture, Branched architecture, and architecture with self-attention modules. Built a 2-stage 3D object detector to detect vehicles in autonomous driving scenes.

Skills and Hobbies