M.Sc. Stefan Schubert
My research interests are within the field of mobile robotics and artificial intelligence. I am working on environment perception, navigation, and machine learning for decision making and object detection.
More information can be found here:
- Exploiting Structural Knowledge for Place Recognition
- Neurologically inspired Place Recognition and Localization
- Circular Convolutional Neural Networks (CCNNs)
- Vector Symbolic Architectures
- Camera-based Navigation in a 3D Map
RSS Pioneer 2019 (Robotics: Science and Systems)
I was one of the 23 international PhD students and postdocs who have been accepted to this year's RSS Pioneers Workshop (RSSP'19). Thanks to the RSSP'19 general chairs Tesca Fitzgerald (Georgia Institute of Technology, USA) and Abhinav Valada (University of Freiburg, Germany) as well as all other organizers.
TAROS 2017 Best Paper Award
We presented the best paper during the Conference Towards Autonomous Robotic Systems (TAROS) held at the University of Surrey, UK. Special thanks to the conference chair Prof. Yang Gao and all other organizers.
TAROS 2016 Best Paper Award
We presented the best paper during the Conference Towards Autonomous Robotic Systems (TAROS) held at the University of Sheffield, UK. Special thanks to the conference chairs Prof. Tony Prescott and Prof. Jacques Penders and all other organizers.
Prize of the Dresden Circle of Business and Science, 2015
Award winner as junior scientist in the field of natural sciences, Dresden, Germany.
German National Scholarship, 2011-2013
Successful application for the German National Scholarship (merit-based scholarship).
An Introduction to Vector Symbolic Architectures and Hyperdimensional Computing, ECAI 2020
We held a tutorial (tutorial website) on high dimensional computing at the 2020 European Conference on Artificial Intelligence (ECAI). We gave an introduction to properties of the high dimensional space, Vector Symbolic Architectures (VSAs), high-dimensional encoding of real world data, and applications. More information about the topic can be found in our recent journal article:
High dimensional computing - the upside of the curse of dimensionality, KI 2019
We held a tutorial (tutorial website) on high dimensional computing at the 2019 German conference on artificial intelligence (KI) in Kassel, Germany. We gave an introduction to properties of the high dimensional space, Vector Symbolic Architectures (VSAs), high-dimensional encoding of real world data, and applications. More information about the topic can be found in our recent journal article:
Circular Convolutional Neural Networks for Panoramic Images and Laser Data
Abstract: Circular Convolutional Neural Networks (CCNN) are an easy to use alternative to CNNs for input data with wrap-around structure like 360° images and multi-layer laserscans. Although circular convolutions have been used in neural networks before, a detailed description and analysis is still missing. This paper closes this gap by defining circular convolutional and circular transposed convolutional layers as the replacement of their linear counterparts, and by identifying pros and cons of applying CCNNs. We experimentally evaluate their properties using a circular MNIST classification and a Velodyne laserscanner segmentation dataset. For the latter, we replace the convolutional layers in two state-of-the-art networks with the proposed circular convolutional layers. Compared to the standard CNNs, the resulting CCNNs show improved recognition rates in image border areas. This is essential to prevent blind spots in the environmental perception. Further, we present and evaluate how weight transfer can be used to obtain a CCNN from an available, readily trained CNN. Compared to alternative approaches (e.g. input padding), our experiments show benefits of CCNNs and transfered CCNNs regarding simplicity of usage (once the layer implementations are available), performance and runtime for training and inference. Implementations for Keras with Tensorflow are provided online