Refine
Year of publication
Document Type
- Conference Proceeding (1253) (remove)
Conference Type
- Konferenzartikel (950)
- Konferenz-Abstract (156)
- Konferenzband (77)
- Sonstiges (42)
- Konferenz-Poster (32)
Language
- English (934)
- German (317)
- Multiple languages (1)
- Russian (1)
Keywords
- Mikroelektronik (62)
- RoboCup (32)
- Gamification (12)
- Machine Learning (12)
- injury (10)
- Biomechanik (9)
- Finite-Elemente-Methode (9)
- Kommunikation (9)
- Assistive Technology (8)
- Produktion (8)
Institute
- Fakultät Elektrotechnik und Informationstechnik (E+I) (bis 03/2019) (453)
- Fakultät Maschinenbau und Verfahrenstechnik (M+V) (286)
- Fakultät Elektrotechnik, Medizintechnik und Informatik (EMI) (ab 04/2019) (213)
- Fakultät Wirtschaft (W) (164)
- Fakultät Medien und Informationswesen (M+I) (bis 21.04.2021) (120)
- ivESK - Institut für verlässliche Embedded Systems und Kommunikationselektronik (113)
- INES - Institut für nachhaltige Energiesysteme (59)
- IMLA - Institute for Machine Learning and Analytics (46)
- ACI - Affective and Cognitive Institute (40)
- Fakultät Medien (M) (ab 22.04.2021) (33)
- WLRI - Work-Life Robotics Institute (21)
- CRT - Campus Research & Transfer (13)
- IBMS - Institute for Advanced Biomechanics and Motion Studies (ab 16.11.2022) (12)
- POIM - Peter Osypka Institute of Medical Engineering (10)
- Zentrale Einrichtungen (10)
- IUAS - Institute for Unmanned Aerial Systems (9)
Open Access
- Open Access (560)
- Closed Access (456)
- Closed (223)
- Bronze (214)
- Diamond (29)
- Grün (13)
- Gold (6)
- Hybrid (6)
Multiple Object Tracking (MOT) is a long-standing task in computer vision. Current approaches based on the tracking by detection paradigm either require some sort of domain knowledge or supervision to associate data correctly into tracks. In this work, we present a self-supervised multiple object tracking approach based on visual features and minimum cost lifted multicuts. Our method is based on straight-forward spatio-temporal cues that can be extracted from neighboring frames in an image sequences without supervision. Clustering based on these cues enables us to learn the required appearance invariances for the tracking task at hand and train an AutoEncoder to generate suitable latent representations. Thus, the resulting latent representations can serve as robust appearance cues for tracking even over large temporal distances where no reliable spatio-temporal features can be extracted. We show that, despite being trained without using the provided annotations, our model provides competitive results on the challenging MOT Benchmark for pedestrian tracking.
Diffracted waves carry high resolution information that can help interpreting fine structural details at a scale smaller than the seismic wavelength. Because of the low signal-to-noise ratio of diffracted waves, it is challenging to preserve them during processing and to identify them in the final data. It is, therefore, a traditional approach to pick manually the diffractions. However, such task is tedious and often prohibitive, thus, current attention is given to domain adaptation. Those methods aim to transfer knowledge from a labeled domain to train the model, and then infer on the real unlabeled data. In this regard, it is common practice to create a synthetic labeled training dataset, followed by testing on unlabeled real data. Unfortunately, such procedure may fail due to the existing gap between the synthetic and the real distribution since quite often synthetic data oversimplifies the problem, and consequently the transfer learning becomes a hard and non-trivial procedure. Furthermore, deep neural networks are characterized by their high sensitivity towards cross-domain distribution shift. In this work, we present deep learning model that builds a bridge between both distributions creating a semi-synthetic datatset that fills in the gap between synthetic and real domains. More specifically, our proposal is a feed-forward, fully convolutional neural network for imageto-image translation that allows to insert synthetic diffractions while preserving the original reflection signal. A series of experiments validate that our approach produces convincing seismic data containing the desired synthetic diffractions.
This paper describes a comparative study of two tactile systems supporting navigation for persons with little or no visual and auditory perception. The efficacy of a tactile head-mounted device (HMD) was compared to that of a wearable device, a tactile belt. A study with twenty participants showed that the participants took significantly less time to complete a course when navigating with the HMD, as compared to the belt.
Machine learning (ML) has become highly relevant in applications across all industries, and specialists in the field are sought urgently. As it is a highly interdisciplinary field, requiring knowledge in computer science, statistics and the relevant application domain, experts are hard to find. Large corporations can sweep the job market by offering high salaries, which makes the situation for small and medium enterprises (SME) even worse, as they usually lack the capacities both for attracting specialists and for qualifying their own personnel. In order to meet the enormous demand in ML specialists, universities now teach ML in specifically designed degree programs as well as within established programs in science and engineering. While the teaching almost always uses practical examples, these are somewhat artificial or outdated, as real data from real companies is usually not available. The approach reported in this contribution aims to tackle the above challenges in an integrated course, combining three independent aspects: first, teaching key ML concepts to graduate students from a variety of existing degree programs; second, qualifying working professionals from SME for ML; and third, applying ML to real-world problems faced by those SME. The course was carried out in two trial periods within a government-funded project at a university of applied sciences in south-west Germany. The region is dominated by SME many of which are world leaders in their industries. Participants were students from different graduate programs as well as working professionals from several SME based in the region. The first phase of the course (one semester) consists of the fundamental concepts of ML, such as exploratory data analysis, regression, classification, clustering, and deep learning. In this phase, student participants and working professionals were taught in separate tracks. Students attended regular classes and lab sessions (but were also given access to e-learning materials), whereas the professionals learned exclusively in a flipped classroom scenario: they were given access to e-learning units (video lectures and accompanying quizzes) for preparation, while face-to-face sessions were dominated by lab experiments applying the concepts. Prior to the start of the second phase, participating companies were invited to submit real-world problems that they wanted to solve with the help of ML. The second phase consisted of practical ML projects, each tackling one of the problems and worked on by a mixed team of both students and professionals for the period of one semester. The teams were self-organized in the ways they preferred to work (e.g. remote vs. face-to-face collaboration), but also coached by one of the teaching staff. In several plenary meetings, the teams reported on their status as well as challenges and solutions. In both periods, the course was monitored and extensive surveys were carried out. We report on the findings as well as the lessons learned. For instance, while the program was very well-received, professional participants wished for more detailed coverage of theoretical concepts. A challenge faced by several teams during the second phase was a dropout of student members due to upcoming exams in other subjects.
Short-term load forecasting (STLF) has been playing a key role in the electricity sector for several decades, due to the need for aligning energy generation with the demand and the financial risk connected with forecasting errors. Following the top-down approach, forecasts are calculated for aggregated load profiles, meaning the sum of singular loads from consumers belonging to a balancing group. Due to the emerging flexible loads, there is an increasing relevance for STLF of individual factories. These load profiles are typically more stochastic compared to aggregated ones, which imposes new requirements to forecasting methods and tools with a bottom-up approach. The increasing digitalization in industry with enhanced data availability as well as smart metering are enablers for improved load forecasts. There is a need for STLF tools processing live data with a high temporal resolution in the minute range. Furthermore, behin-the-meter (BTM) data from various sources like submetering and production planning data should be integrated in the models. In this case, STLF is becoming a big data problem so that machine learning (ML) methods are required. The research project “GaIN” investigates the improvement of the STLF quality of an energy utility using BTM data and innovative ML models. This paper describes the project scope, proposes a detailed definition for a benchmark and evaluates the readiness of existing STLF methods to fulfil the described requirements as a reviewing paper.
The review highlights that recent STLF investigations focus on ML methods. Especially hybrid models gain more and more importance. ML can outperform classical methods in terms of automation degree and forecasting accuracy. Nevertheless, the potential for improving forecasting accuracy by the use of ML models depends on the underlying data and the types of input variables. The described methods in the analyzed publications only partially fulfil the tool requirements for STLF on company level. There is still a need to develop suitable ML methods to integrate the expanded data base in order to improve load forecasts on company level.
The interaction between agents in multiagent-based control systems requires peer to peer communication between agents avoiding central control. The sensor nodes represent agents and produce measurement data every time step. The nodes exchange time series data by using the peer to peer network in order to calculate an aggregation function for solving a problem cooperatively. We investigate the aggregation process of averaging data for time series data of nodes in a peer to peer network by using the grouping algorithm of Cichon et al. 2018. Nodes communicate whether data is new and map data values according to their sizes into a histogram. This map message consists of the subintervals and vectors for estimating the node joining and leaving the subinterval. At each time step, the nodes communicate with each other in synchronous rounds to exchange map messages until the network converges to a common map message. The node calculates the average value of time series data produced by all nodes in the network by using the histogram algorithm. The relative error for comparing the output of averaging time series data, and the ground truth of the average value in the network will decrease as the size of the network increases. We perform simulations which show that the approximate histograms method provides a reasonable approximation of time series data.
We propose in this work to solve privacy preserving set relations performed by a third party in an outsourced configuration. We argue that solving the disjointness relation based on Bloom filters is a new contribution in particular by having another layer of privacy on the sets cardinality. We propose to compose the set relations in a slightly different way by applying a keyed hash function. Besides discussing the correctness of the set relations, we analyze how this impacts the privacy of the sets content as well as providing privacy on the sets cardinality. We are in particular interested in how having bits overlapping in the Bloom filters impacts the privacy level of our approach. Finally, we present our results with real-world parameters in two concrete scenarios.
The Effect of Gamification on Emotions - The Potential of Facial Recognition in Work Environmentsns
(2015)
Gamification means using video game elements to improve user experience and user engagement in non-game services and applications. This article describes the effects when gamification is used in work contexts. Here we focus on industrial production. We describe how facial recognition can be employed to measure and quantify the effect of gamification on the users’ emotions.
The quantitative results show that gamification significantly reduces both task completion time and error rate. However, the results concerning the effect on emotions are surprising. Without gamification there are not only more unhappy expressions (as to expect) but surprisingly also more happy expressions. Both findings are statistically highly significant.
We think that in redundant production work there are generally more (negative) emotions involved. When there is no gamification happy and unhappy balance each other. In contrast gamification seems to shift the spectrum of moods towards “relaxed”. Especially for work environments such a calm attitude is a desirable effect on the users. Thus our findings support the use of gamification.
Video game developers continuously increase the degree of details and realism in games to create more human-like characters. But increasing the human-likeness becomes a problem in regard to the Uncanny Valley phenomenon that predicts negative feelings of people towards artificial entities. We developed an avatar creation system to examine preferences towards parametrized faces and explore in regard to the Uncanny Valley phenomenon how people design faces that they like or reject. Based on the 3D model of the Caucasian average face, 420 participants generate 1341 faces of positively and negatively associated concepts of both gender. The results show that some characteristics associated with the Uncanny Valley are used to create villains or repulsive faces. Heroic faces get attractive features but are rarely and little stylized. A voluntarily designed face is very similar to the heroine. This indicates that there is a tendency of users to design feminine and attractive but still credible faces.
The precise positioning of mobile systems is a prerequisite for any autonomous behavior, in an industrial environment as well as for field robotics. The paper describes the set up for an experimental platform and its use for the evaluation of simultaneous localization and mapping (SLAM) algorithms. Two approaches are compared. First, a local method based on point cloud matching and integration of inertial measurement units is evaluated. Subsequent matching makes it possible to create a three-dimensional point cloud that can be used as a map in subsequent runs. The second approach is a full SLAM algorithm, based on graph relaxation models, incorporating the full sensor suite of odometry, inertial sensors, and 3D laser scan data.