3DUU
3D Urban Understanding
Through developments in 3D sensor technology, photogrammetry and computer vision, real-world urban scenes can now be captured at a large scale in the form of images or point clouds. This data could support powerful models for applications such as urban planning and self-driving vehicles. However, robustly and efficiently representing these large and dynamic outdoor scenes in a useable semantic 3D representation remains challenging.
In the 3DUU Lab, we will develop new methods and techniques that automatically recognise and model objects in real-world scenes in 3D by combining data from various sources, such as aerial photos and laser scanners on vehicles. We investigate localizing 2D images in the 3D world, reconstructing 3D scenes from such images, and subsequently recognizing objects from 3D or even from multiple sensing modalities simultaneously. Our techniques can thus enrich the data with information about the location and types of objects and surfaces in the scene, such as buildings, streets, trees, traffic lights, and terrains.
The 3DUU Lab is part of the TU Delft AI Labs programme.
Challenges in large-scale urban scene reconstruction from images
Real-time 3D sensing for autonomous driving using a roof-mounted LiDAR
Being able to control the flow around airfoils can yield huge benefits in terms of acoustic emission or drag reduction. AI techniques will play a key role in achieving this by identifying novel active flow control strategies or efficiently designing the geometry of porous airfoil trailing edge.
The team
Directors
Liangliang Nan
Faculty of ABE
Lab director
Julian Kooij
Faculty of ME
Lab director
PhD candidates
Nail Ibrahimli
PhD candidate
Shenglan Du
PhD candidate
Mubariz Zaffar
PhD candidate
Shiming Wang
PhD candidate
Education
Courses
2024/2025
- Machine Learning for the Built Environment | GEO5017
- Machine Learning for Robotics | RO47002
- Robot Software Practical | RO47003
- Photogrammetry and 3D Computer Vision | GEO1016
2023/2024
- Machine Learning for the Built Environment | GEO5017
- Machine Learning for Robotics | RO47002
- Robot Software Practical | RO47003
- Photogrammetry and 3D Computer Vision | GEO1016
2022/2023
- Machine Learning for the Built Environment | GEO5017
- Machine Learning for Robotics | RO47002
- Robot Software Practical | RO47003
- Photogrammetry and 3D Computer Vision | GEO1016
2021/2022
- Machine Learning for the Built Environment | GEO5017
- Machine Learning for Robotics | RO47002
- Robot Software Practical | RO47003
- Photogrammetry and 3D Computer Vision | GEO1016
2020/2021
- Robot Software Practical | RO47003
- Photogrammetry and 3D Computer Vision | GEO1016
2019/2020
- Photogrammetry and 3D Computer Vision | GEO1016
Master Projects
Openings
- Accurate Robot Localization using only Camera Images
Ongoing
- Acoustic Traffic Perception, Julian Kooij, Boriss Bermans (2023/2024)
- BEP project: nuScenes light – Compressing an autonomous driving dataset for efficient deep learning, Julian Kooij, Tim Bergervoet (2023/2024)
- BEP project: nuScenes light – Compressing an autonomous driving dataset for efficient deep learning, Julian Kooij, Friso de Swart (2023/2024)
- BEP project: nuScenes light – Compressing an autonomous driving dataset for efficient deep learning, Julian Kooij, Willem Vromans (2023/2024)
- BEP project: nuScenes light – Compressing an autonomous driving dataset for efficient deep learning, Julian Kooij, Liam Punselie (2023/2024)
- A deep learning approach to improve the classification of airborne lidar point clouds, Liangliang Nan, Sharath Chandra Madanu (2023/2024)
- Multi view diffusion for geometric tasks, Liangliang Nan, Chi Zhang (2023/2024)
- Enhancing 3D City Models with Neural Representations, Liangliang Nan, Sitong Li (2023/2024)
- Text-guided geometry editting, Liangliang Nan, Yingxin Feng (2023/2024)
- Centralized benchmark for 3D vision tasks, Liangliang Nan, Serenic Monte (2023/2024)
- Procedural Modelling of Tree Growth Using Multi-temporal Point Clouds, Liangliang Nan, Noortje van der Horst (2021/2022)
- Accurate Robot Localization using only Camera Images, Julian Kooij (2023/2024)
Finished
- BEP project: Visual Localization in Delft for where GPS fails, Julian Kooij, Sisko van Munster (2023/2024)
- BEP project: Visual Localization in Delft for where GPS fails, Julian Kooij, Jimmie Kwok (2023/2024)
- BEP project: Visual Localization in Delft for where GPS fails, Julian Kooij, Stan Marseille (2023/2024)
- Domain shift-aware Ensemble-based Visual Place Recognition, Julian Kooij, Wouter de Leeuw (2022/2023)
- Using ensembles of Visual Place Recognition techniques for vehicle localisation, Julian Kooij, Marios Marinos (2022/2023)
- Using ensembles of Visual Place Recognition techniques for vehicle localisation, Julian Kooij, Ruben Sangers (2022/2023)
- Using ensembles of Visual Place Recognition techniques for vehicle localisation, Julian Kooij, Pol Mur i Uribe (2022/2023)
- Can Neural Radiance Fields (NeRF) reconstruct a street from a self-driving vehicle's camera images? Julian Kooij, Alexander Freeman (2022/2023)
- Can Neural Radiance Fields (NeRF) reconstruct a street from a self-driving vehicle's camera images? Julian Kooij, Alessandro Duico (2022/2023)
- Can Neural Radiance Fields (NeRF) reconstruct a street from a self-driving vehicle's camera images? Julian Kooij, Boriss Bermans (2022/2023)
- Floor count from street view imagery using learning-based façade parsing, Liangliang Nan, Daniel Dobson (2022/2023)
- Neural Surface Reconstruction and Stylization, Liangliang Nan, Fabian Visser (2022/2023)
- Multi View Semantic Reconstruction, Liangliang Nan, Ioanna Panagiotidou (2022/2023)
- Shape-guided artistic route finding, Liangliang Nan, Leon Powalka (2022/2023)
- Detailed Facade Reconstruction for Mahattan-world Buildings, Liangliang Nan, Linjun Wang (2021/2022)
- Automated Semantic Segmentation of Aerial Imagery using Synthetic Data, Liangliang Nan, Camilo Caceres Tocora (2021/2022)
- Learning to Reconstruct Compact Building Models from Point Clouds, Liangliang Nan, Zhaiyu Chen (2020/2021)
