Cedric Perauer

cedric@CV: ~
Cedric

User: Cedric Perauer

OS: Computer Vision/ML Engineer @ Hexagon

TUM logo Hexagon logo CombatIQ logo Apple logo Bosch logo Mercedes-Benz logo
cedric@CV: ~
[cedric] 0:about 1:publications 2:projects 3:patents

cedric@CV:~$cat welcome.txt

Hello! I'm Cedric :wave: I work as a computer vision/ML engineer at Hexagon, and I am interested in scene understanding and generation.

Previously, I was a master's student at TUM and did my thesis with Dr. Alexey Artemov at the TUM Visual Computing & AI Lab. I also worked on computer vision for robotics at Apple, automated driving at Bosch and Mercedes-Benz, and was a founding computer vision engineer at CombatIQ, where I shipped real-time computer vision models for UFC fights.

I am passionate about bringing the latest research into products and working on the full stack, from product-driven research to runtime optimization and deployment of models. I enjoy reimplementing papers and following the latest research, both at work and in my free time.

cedric@CV:~$/selected-publications

  1. autoinst.jpg
    AutoInst: Automatic Instance-Based Segmentation of LiDAR 3D Scans
    Cedric Perauer, Laurenz Adrian Heidrich, Haifan Zhang, Matthias Nießner, Anastasiia Kornilova*, and Alexey Artemov*
    In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024
    arXiv Project Page Code

cedric@CV:~$/selected-projects

  1. project thumbnail
    Reimplementation of LoL: Longer than Longer, Scaling Video Generation to Hour
    An implementation of the 2026 Bytedance & UCLA paper LoL
    Project Page Code
  2. project thumbnail
    Instance Deformable Metal Splatter
    Instance aware real-time dynamic 3D gaussian splatting on Apple devices, based on my friend's (Li et al.) paper TRASE [3DV 2026]
    Project Page Code
  3. project thumbnail
    Latent Generative Anamorphoses
    2D latent generative anamorphoses with Stable Diffusion 3.5, inspired by LookingGlass (CVPR2025)
    Project Page Code
  4. project thumbnail
    Training-Free 360° Panorama Generation
    Progressive inpainting via equirectangular projection based on Meta's 2025 ICCV paper "A recipe for generating 3D worlds from a single image"
    Project Page Code
  5. project thumbnail
    Nvidia SANA Inpainting
    Inpainting implementation for Nvidia SANA [ICLR2025] using masked Flow scheduler
    Project Page Code

cedric@CV:~$/selected-patents

Method for determining a driving behaviour and method for adapting control algorithms of automated driving systems

US12221114B2 June 2025

We compare a human driver's manual driving behaviour against an automated driving system and use the difference to adapt the system's control algorithms toward more human-like driving.

Method and apparatus for identifying a parking space for a vehicle by means of machine learning methods

WO2023237549A1 December 2023

We propose a novel backbone agnostic neural network head that builds on the geometrical characteristics of parking spaces.

Process for camera-based environment detection

DE102021002798A1 July 2021

Our method proposes the use of Neural Rendering (NeRF) for environment perception in the ADAS space.

  1. fsd1.png
    Development and Deployment of a Perception Stack for the Formula Student Driverless Competition
    Cedric Perauer
    2021
    ResearchGate Code (partial)
  2. autoinst.jpg
    AutoInst: Automatic Instance-Based Segmentation of LiDAR 3D Scans
    Cedric Perauer, Laurenz Adrian Heidrich, Haifan Zhang, Matthias Nießner, Anastasiia Kornilova*, and Alexey Artemov*
    In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024
    arXiv Project Page Code
  1. project thumbnail
    Reimplementation of LoL: Longer than Longer, Scaling Video Generation to Hour
    An implementation of the 2026 Bytedance & UCLA paper LoL
    Project Page Code
  2. project thumbnail
    Instance Deformable Metal Splatter
    Instance aware real-time dynamic 3D gaussian splatting on Apple devices, based on my friend's (Li et al.) paper TRASE [3DV 2026]
    Project Page Code
  3. project thumbnail
    Latent Generative Anamorphoses
    2D latent generative anamorphoses with Stable Diffusion 3.5, inspired by LookingGlass (CVPR2025)
    Project Page Code
  4. project thumbnail
    Training-Free 360° Panorama Generation
    Progressive inpainting via equirectangular projection based on Meta's 2025 ICCV paper "A recipe for generating 3D worlds from a single image"
    Project Page Code
  5. project thumbnail
    A Simple Virtual Staging Pipeline
    A simple virtual staging pipeline based on Stable Diffusion trained in Instruct Pix2Pix fashion with a paired image dataset
    Project Page
  6. project thumbnail
    Dynamic NeRF Methods Comparison
    A comparison study of various dynamic NeRF methods
    Project Page Code
  7. project thumbnail
    Few Shot Robot Grasping with Dinov2 Backbone
    A method for data efficient top down robot grasp detection with Dinov2
    Project Page Code
  8. project thumbnail
    Nvidia SANA Inpainting
    Inpainting implementation for Nvidia SANA [ICLR2025] using masked Flow scheduler
    Project Page Code

Method for determining a driving behaviour and method for adapting control algorithms of automated driving systems

US12221114B2 June 2025

We compare a human driver's manual driving behaviour against an automated driving system and use the difference to adapt the system's control algorithms toward more human-like driving.

Method and apparatus for identifying a parking space for a vehicle by means of machine learning methods

WO2023237549A1 December 2023

We propose a novel backbone agnostic neural network head that builds on the geometrical characteristics of parking spaces.

Detector for detecting vibration of an object and method

DE102021003560A1 January 2023

We measure an object's vibrations remotely by projecting a laser interference pattern onto it and evaluating the motion of the interference fringes.

Method for training an autonomous driving function

DE102021004426A1 November 2021

We record environment, driving-dynamics, and driver-command data from drivers classified as safe and use it to train an autonomous driving function by imitation.

Method for user evaluation, control device for carrying out such a method, and vehicle

DE102021002301A1 July 2021

We capture the occupant's eyes with an in-cabin sensor and reconstruct their gaze path to evaluate the user's attention.

Process for camera-based environment detection

DE102021002798A1 July 2021

Our method proposes the use of Neural Rendering (NeRF) for environment perception in the ADAS space.

Process for the formation of a vehicle group

DE102021002160A1 June 2021

We connect vehicles whose individual automation level is insufficient into a coupled vehicle group that can operate automated or driverless together.

Procedure for operating vehicles within a platooning convoy

DE102021002004A1 June 2021

All vehicles of a platooning convoy are registered in a backend, which centrally establishes the driving order of the platoon.

Method of operating a vehicle

DE102021000509A1 March 2021

We segment the occupant's body pose from stereo-camera images with a pose segmentation model and derive their optimal seating position from it.