Student Research Assistant in 3D Semantic Building - Munich, Deutschland - Technische Universität München

Technische Universität München

Geprüftes Unternehmen

Munich, Deutschland

vor 2 Wochen

Geschrieben von:

Lena Wagner

beBee Recruiter

Beschreibung

Creating detailed semantic 3D building models is a long
- standing challenge in photogrammetry and computer vision. While methods

utilizing 2D building footprints and aerial imagery have enabled generating

models up to Level of Detail (LoD) 2, robust solutions achieving facade
- detailed semantic LoD3 models remain in infancy [1, 2].

Automated reconstruction of LoD3-specific facade elements is crucial for

or simulate floods. Our studies [1] suggest that mobile mapping data, par
- ticularly from units such as Google Street View, holds great potential for

semantic LoD3 facade modeling. However, utilizing such data requires ro
- bust automatic image projection, semantic segmentation, and reconstruction

algorithms.

Objectives In this project, you will be supervised by Olaf Wysocki and

work closely with Prof Biljecki's group of the National University of Singa
- pore. This position is for 8 hours/week, starting in April/May until the end

of the semester, September/October, with the possibility of extension; The

joint publication is foreseen

Your tasks will include:

Acquiring automatically highquality streetlevel images from crowd

- sourcing platforms; providing ground-truth data

Adapting existing image projection methods for the streetview image

- to-model scenario

Testing and finetuning of the stateoftheart methods for facade se

- mantic segmentation

Analyzing the impact of image segmentation on 3D semantic recon

- struction

Requirements Sound programming skills in Python are required. Fa
- miliarity with libraries such as OpenCV, Open3D, and PyTorch is highly

advantageous. Understanding machine learning, computer vision, and pho
- togrammetry fundamentals is required. Student status, CV, and the current

References [1] Wysocki, O., Xia, Y., Wysocki M., Grilli, E., Hoegner,

L., Cremers D., and Stilla, U.

Scan2LoD3:
Reconstructing semantic 3D

building models at LoD3 using ray casting and Bayesian networks, CVPRW