Lehrstuhl für Allgemeine Elektrotechnik und Theoretische Nachrichtentechnik
Prof. Dr.- Ing. Anton Kummert
Drittmittelstipendium zum Thema “Machine-Learning based Trajectory Prediction“
In der Fakultät 6 ist zum 01.03.2020 befristet auf 3 Jahre ein Drittmittelstipendium zum Thema ... [mehr]
Drittmittelstipendium zum Thema “Semi-Automated Ground Truthing of Automotive Sensors“
In der Fakultät 6 ist zum 01.02.2020 befristet auf 3 Jahre ein Drittmittelstipendium zum Thema... [mehr]
Drittmittelstipendium zum Thema “Machine Learning for Radar Signal Processing“
In der Fakultät 6 ist zum 01.01.2020 befristet auf 3 Jahre ein Drittmittelstipendium zum Thema ... [mehr]
Prof. Dr.-Ing. Anton Kummert
|Telefon:||+49 (0)202 439 1961|
Fast and Reliable Architecture Selection for Convolutional Neural Networks Proceedings of the 27th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN), pages 179-184, Bruges 2019ArXiv, abs/1905.01924
Schlüsselwörter: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
Zusammenfassung: The performance of a Convolutional Neural Network (CNN) depends on its hyperparameters, like the number of layers, kernel sizes, or the learning rate for example. Especially in smaller networks and applications with limited computational resources, optimisation is key. We present a fast and efficient approach for CNN architecture selection. Taking into account time consumption, precision and robustness, we develop a heuristic to quickly and reliably assess a network's performance. In combination with Bayesian optimisation (BO), to effectively cover the vast parameter space, our contribution offers a plain and powerful architecture search for this machine learning technique.
False Alarm Reduction for Active Sonars using Deep Learning Architectures
2019 UDT, Stockholm
Driver State Monitoring with Hierarchical Classification
2018 IEEE International Conference on Intelligent Transportation Systems (ITSC)
Two-dimensional (2D) systems approach to feedforward/feedback control of a class of spatially interconnected systemsInternational Journal of Control, 91:1-23
Zusammenfassung: Electrical ladder circuits, consisting of a series, or cascade, connection of cells are a class of spatially interconnected systems. These circuits can be modeled as 2D systems, i.e., there exist two directions of information propagation, where one indeterminate is time and the other the number of the current cell (node). In this paper, the recently developed direct (2D) approach to stability analysis and stabilization of these systems is extended to the presence of uncertainty in the models described by the norm bounded structure. The analysis is then further extended to the design of feedforward/feedback control action to track a spatially distributed time invariant reference signal in the presence of disturbances.
Dense Spatial Translation Network
2018 IEEE International Conference on Vehicular Electronics and Safety (ICVES)
Evaluation un terschiedlicher Klassifikationsalgorithmen zur Falschalarmreduktion in der Aktiv-Sonarortung
Jahrestagung für Akustik DAGA 2018
Conditional multichannel generative adversarial networks with an application to traffic signs representation learningProgress in Artificial Intelligence, 8
Zusammenfassung: Generative adversarial networks (GANs) are known to produce photorealistic representations. However, we show in this study that this is only valid when the input channels come from a regular RGB camera sensor. In order to alleviate this shortcoming, we propose a general solution to which we refer to as multichannel GANs (MCGANs). In contrast to the existing approaches, MCGANs can process multiple channels with different textures and resolutions. This is achieved by using known concepts in deep learning such as weight sharing and specially separated convolutions. The proposed pipeline enables particular kernels to learn low-level characteristics from the different channels without the need for exhaustive hyper-parameter tuning. We demonstrate the improved representational ability of the framework on traffic sign samples that are captured by a camera with a so-called red-clear-clear-clear pixel topology. Furthermore, we extend our solution by applying the concept of conditions, that offers a whole spectrum of new features, especially for the generation of traffic signs. Throughout this paper, we further discuss relevant applications for the generated synthetic data.
Aggregated channels network for real-time pedestrian detection
, Seite 54.
Recent Advances in Machine Learning for Signal Analysis and ProcessingJournal of The Franklin Institute, Special Issue, 355(4):1513-2066
A k-D Stability Measure for Discrete Roesser-Like System Implementations
Feature investigation and control systems design for spatially interconnected systems
Two-dimensional(2D) systems approarch to feedforward/feedback control of a classofspatially interconnected systems