004 Informatik
Refine
Document Type
- Article (10)
- Working Paper (2)
- Bachelor Thesis (1)
- Book (1)
- Conference Proceeding (1)
- Study Thesis (1)
Keywords
- Neuronales Netz (3)
- Data Mining (2)
- Kalman-Filter (2)
- data analysis (2)
- data mining (2)
- model car (2)
- neural networks (2)
- training data (2)
- virtual (2)
- Ableitung Solltrajektorie (1)
OAFuser: Online Adaptive Extended Object Tracking and Fusion using automotive Radar Detections
(2020)
This paper presents the Online Adaptive Fuser: OAFuser, a novel method for online adaptive estimation of motion and measurement uncertainties for efficient tracking and fusion by applying a system of several estimators for ongoing noise along with the conventional state and state covariance estimation. In our system, process and measurement noises are estimated with steady-state filters to obtain combined measurement noise and process noise estimators for all sensors in order to obtain state estimation with a linear Minimum Mean Square Error (MMSE) estimator and accelerating the system’s performance. The proposed adaptive tracking and fusion system was tested based on high fidelity simulation data and several real-world scenarios for automotive radar, where ground truth data is available for evaluation. We demonstrate the proposed method’s accuracy and efficiency in a challenging, highly dynamic scenario where our system is benchmarked with Multiple Model filter in terms of error statistics and run time performance.
In recent years, machine learning methods have taken a
firm place in society and their use continues to grow. The challenge
here is their little to almost non-existent interpretability. The aim of this
paper is to uncover the possibilities of interpreting machine learning. The
novel mechanisms and procedures of the emerging field of interpretable
machine learning are presented. In a two-part analysis, intrinsically
interpretable machine learning methods and established post-hoc interpretation
methods are examined in more detail. The focus is on their
functionality, properties and boundary conditions. Finally, a use case
will be used as an example to demonstrate how post-hoc interpretation
methods can contribute to the explainability of an image classifier and
systematically provide new insights into a model.
A matter of reality
(2018)
Due to the increasing relevance of data, more and more data from various sources is accumulated for a variety of purposes. At the same time, however, there is a shortage of data in areas where it is urgently needed. Particularly in the field of machine learning, there is a lack of good and usable training data. Therefore, this research paper is concerned with the virtual data acquisition for the training of neural networks. For this purpose, first an application was developed that aims to generate virtual, automatically labeled data. Subsequently, a neural network was trained on the generated virtual data and tested on real data.
A matter of reality
(2018)
Due to the increasing relevance of data, more and more data from various sources is accumulated for a variety of purposes. At the same time, however, there is a shortage of data in areas where it is urgently needed. Particularly in the field of machine learning, there is a lack of good and usable training data. Therefore, this research paper is concerned with the virtual data acquisition for the training of neural networks. For this purpose, first an application was developed that aims to generate virtual, automatically labeled data. Subsequently, a neural network was trained on the generated virtual data and tested on real data.
Diese Arbeit beschreibt ein Referenzsystem,
welches die Trajektorie eines Modellfahrzeugs mithilfe einer
Deckenkamera ermittelt.
Hierbei wird die Position eines Fahrzeugs durch die Kombination
von mehreren Bildverarbeitungsverfahren bestimmt.
Anschließend erfolgt die Verbesserung der Messdaten sowie die
Schätzung von nicht direkt messbaren Größen mithilfe eines
Partikelfilters. Die abschließende Zuordnung der Referenzinformationen
zu den On-Board-Messwerten wird durch eine
Zeitsynchronisation zwischen Fahrzeug und Referenzsystem
ermöglicht.
Das entwickelte System stellt somit eine hochgenaue Vergleichsbasis
für die Validation und die Abschätzung der
Genauigkeit von Lokalisationsverfahren bereit und erleichtert
daher die Entwicklung autonomer Fahralgorithmen.
Diese Arbeit zeigt Wege und Überlegungen zur Entwicklung eines räumlichen Interakti- onskonzeptes auf. Als Basis diente die Vorgabe eines Eingabegerätes, der Fernbedienung der Spielekonsole Wii von Nintendo. Die Wii-Fernbedienung, auch Wiimote genannt, dient dem Benutzer einer Wii-Konsole als primäres Spiel- und Steuerinstrument. Bisherige Benutzerschnittstellen „fesseln“ den Nutzer oft an eine Sitzgelegenheit. So würde sich beispielsweise niemand auf Dauer mit einer Tastatur in der Hand durch den Raum bewegen. Was wäre also, wenn die Bewegung bereits teil der Eingabe wäre? Die Wiimote hat mit ihren eingebauten Bewegungssensoren diese Bedienungsart einem breitem Publi- kum zugänglich gemacht, so dass der erste Schritt bereits gemacht sein dürfte, die eigenen motorischen Fähigkeiten als Eingabemethode zu nutzen.