Timo Lüddecke

Research Group Leader
Institute of Computer Science
University of Göttingen

eMail Google Scholar Github

My research focuses on scalable, multi-modal machine learning methods that make use of large datasets and use cross-modal relationships as a learning signal. A goal of my research is to minimize effort and facilitate labelling. To this end, data-centric and user interaction techniques play a central role. Current applications involve animal behavior analysis and acoustics engineering.

Group Members

Jan van Delden (PhD student)
Felix B. Müller (PhD student)
Julius Schultz (associated PhD student at TU Braunschweig)
Christian Libner (student assistant)
You?

If you are intersted in joining my group, please write an email directly to me.

Short Bio

Before starting my research group in January 2024, I worked as a postdoc with Alexander Ecker. I completed my PhD in 2019 with Florentin Wörgötter working on robotic perception at the University of Göttingen. I received Bachelor and Master from the University of Magdeburg with an emphasis on medical image processing working with Klaus Tönnies.

Selected Publications

See Google Scholar for an up-to-date and complete list of publications.

Learning to Predict Structural Vibrations

Jan van Delden, Julius Schultz, Christopher Blech, Sabine C Langer, and Timo Lüddecke

accepted for Conference on Neural Information Processing Systems (NeurIPS), 2024

Computer Vision for Primate Behavior Analysis in the Wild

Richard Vogg, Timo Lüddecke, Jonathan Henrich, Sharmita Dey, Matthias Nuske, Valentin Hassler, Derek Murphy, Julia Fischer, Julia Ostner, Oliver Schülke, and others

accepted in principle in Nature Methods, 2024

GPT4GEO: How a Language Model Sees the World's Geography

Jonathan Roberts, Timo Lüddecke, Sowmen Das, Kai Han, and Samuel Albanie

NeurIPS 2023 Workshop on Foundation Models for Decision Making, 2023

Self-supervised graph representation learning for neuronal morphologies

Marissa A Weis, Laura Pede, Timo Lüddecke, and Alexander S Ecker

Transactions on Machine Learning Research (TMLR), 2023

Image Segmentation Using Text and Image Prompts

Timo Lüddecke, and Alexander Ecker

Proceedings of the Conference on Computer Vision and Pattern Recognition (CVPR), 2022

One-shot multi-path planning for robotic applications using fully convolutional networks

Tomas Kulvicius, Sebastian Herzog, Timo Lüddecke, Minija Tamosiunaite, and Florentin Wörgötter

In Proceedings of International Conference on Robotics and Automation (ICRA), 2020

Fine-grained action plausibility rating

Timo Lüddecke, and Florentin Wörgötter

Robotics and Autonomous Systems (RAS), 2020

CNNs can efficiently trace paths, too

Timo Lüddecke, and Alexander Ecker

NeurIPS 2020 Workshop SVRHM, 2020

Distributional Semantics of Objects in Visual Scenes in Comparison to Text

Timo Lüddecke, Alejandro Agostini, Michael Fauth, Minija Tamosiunaite, and Florentin Wörgötter

Artificial Intelligence, 2019

Context-based Affordance Segmentation from 2D Images for Robot Action

Timo Lüddecke, Tomas Kulvicius, and Florentin Wörgötter

Robotics and Autonomous Systems (RAS), 2019

Learning to Segment Affordances

Timo Lüddecke, and Florentin Wörgötter

International Conference on Computer Vision Workshops (ICCVW), 2017

Convolutional Neural Networks for Movement Prediction in Videos

Alexander Warnecke, Timo Lüddecke, and Florentin Wörgötter

German Conference on Pattern Recognition (GCPR), 2017

Action-oriented Scene Understanding

Timo Lüddecke

PhD Thesis

Open Source Software

DrawCNN

DrawCNN is a python script to visualize CNN architectures and export to SVG for later refinement.

Tralo

Tralo is a dependency-free training and experiment framework for PyTorch.