Informatik
Refine
Document Type
- Conference Proceeding (438)
- Article (106)
- Part of a Book (51)
- Doctoral Thesis (11)
- Book (9)
- Anthology (7)
- Patent (2)
- Working Paper (2)
- Report (1)
Is part of the Bibliography
- yes (627)
Institute
- Informatik (627)
- Technik (1)
Publisher
- Springer (110)
- Hochschule Reutlingen (96)
- IEEE (65)
- Gesellschaft für Informatik (58)
- Elsevier (24)
- ACM (17)
- Springer Gabler (15)
- RWTH Aachen (9)
- Università Politecnica delle Marche (8)
- Haufe (7)
Uncontrolled movements of laparoscopic instruments can lead to inadvertent injury of adjacent structures. The risk becomes evident when the dissecting instrument is located outside the field of view of the laparoscopic camera. Technical solutions to ensure patient safety are appreciated. The present work evaluated the feasibility of an automated binary classification of laparoscopic image data using Convolutional Neural Networks (CNN) to determine whether the dissecting instrument is located within the laparoscopic image section. A unique record of images was generated from six laparoscopic cholecystectomies in a surgical training environment to configure and train The CNN. By using a temporary version of the neural network, the annotation of the training image files could be automated and accelerated. A combination of oversampling and selective data augmentation was used to enlarge the fully labelled image data set and prevent loss of accuracy due to imbalanced class volumes. Subsequently the same approach was applied to the comprehensive, fully annotated Cholec80 database. The described process led to the generation of extensive and balanced training image data sets. The performance of the CNN-based binary classifiers was evaluated on separate test records from both databases. On our recorded data, an accuracy of 0.88 with regard to the safety-relevant classification was achieved. The subsequent evaluation on the Cholec80 data set yielded an accuracy of 0.84. The presented results demonstrate the feasibility of a binary classification of laparoscopic image data for the detection of adverse events in a surgical training environment using a specifically configured CNN architecture.
The cloud evolved into an attractive execution environment for parallel applications, which make use of compute resources to speed up the computation of large problems in science and industry. Whereas Infrastructure as a Service (IaaS) offerings have been commonly employed, more recently, serverless computing emerged as a novel cloud computing paradigm with the goal of freeing developers from resource management issues. However, as of today, serverless computing platforms are mainly used to process computations triggered by events or user requests that can be executed independently of each other and benefit from on-demand and elastic compute resources as well as per-function billing. In this work, we discuss how to employ serverless computing platforms to operate parallel applications. We specifically focus on the class of parallel task farming applications and introduce a novel approach to free developers from both parallelism and resource management issues. Our approach includes a proactive elasticity controller that adapts the physical parallelism per application run according to user-defined goals. Specifically, we show how to consider a user-defined execution time limit after which the result of the computation needs to be present while minimizing the associated monetary costs. To evaluate our concepts, we present a prototypical elastic parallel system architecture for self-tuning serverless task farming and implement two applications based on our framework. Moreover, we report on performance measurements for both applications as well as the prediction accuracy of the proposed proactive elasticity control mechanism and discuss our key findings.
Unternehmen wenden insbesondere bei IT-nahen Projekten seit einigen Jahren auch im Controlling verstärkt ein agiles Vorgehen an. Erfahrungen zeigen jedoch, dass dies nicht bei allen Projekten in jedem Unternehmen funktioniert. Hybride Ansätze, die agile mit klassischen Projekt-Management-Methoden verbinden, bieten eine Lösung.
Automatic anode rod inspection in aluminum smelters using deep-learning techniques: a case study
(2020)
Automatic fault detection using machine learning has become an exciting and promising area of research. This because it accurate and timely way to manage and classify with minimal human effort. In the computer vision community, deep-learning methods have become the most suitable approaches for this task. Anodes are large carbon blocks that are used to conduct electricity during the aluminum reduction process. The most basic function of anode rod inspection is to prevent a situation where the anode rod will not fit into the stub-holes of a new anode. It would be the case for a rod containing either severe toe-in, missing stubs, or a retained thimble on one or more stubs. In this work, to improve the accuracy of shape defect inspection for an anode rod, we use the Fast Region-based Convolutional Network method (Fast R-CNN), model. To train the detection model, we collect an image dataset composed of multi-class of anode rod defects with annotated labels. Our model is trained using a small number of samples, an essential requirement in the industry where the number of available defective samples is limited. It can simultaneously detect multi-class of defects of the anode rod in nearly real-time.
Ever since the 1980s, researchers in computer science and robotics have been working on making autonomous cars. Due to recent breakthroughs in research and devel- opment, such as the Bertha Benz Project [ZBS+14], the goal of fully autonomous vehicles seems closer than ever before. Yet a lot of questions remain unanswered. Especially now that the automotive industry moves towards autonomous systems in series production vehicles, the task of precise localization has to be solved with automotive grade sensors and keep memory and processing consumption at a mini- mum. This thesis investigates the Simultaneous Localization and Mapping (SLAM) prob- lem for autonomous driving scenarios on a parking lot using low cost automotive sensors. The main focus is herby devoted to the RAdio Detection And Ranging (RADAR) sensor, which has not been widely analyzed in an autonomous driving scenario so far, even though they are abundant in the automotive industry for ap- plications such as Adaptive Cruise Control (ACC). Due to the high noise floor, the radar sensor has widely been disregarded in the Intelligent Transportation Systems and Robotics communities with regards to SLAM applications. However in this thesis, it is shown that the RADAR sensor proves to be an affordable, robust and precise sensor, when modeling its physical properties correctly. In this regard, a GraphSLAM based framework is introduced, which extracts features from the RADAR sensor and generates an optimized map of the surroundings using the RADAR sensor alone. This framework is used to enable crowd based localization, which is not limited to the RADAR sensor alone. By integrating an automotive Light Detection and Ranging (LiDAR) and stereo camera sensor, a robust and precise localization system can be built that that is suitable for autonomous driving even in complex parking lot scenarios. It it is thereby shown that the RADAR sensor is strongly contributing to obtaining good results in a sensor fusion setup. These results were obtained on an extensive dataset on a parking lot, which has been recorded over the course of several months. It contains different weather conditions, different configurations of parked cars and a multitude of different trajectories to validate the approaches described in this thesis and to come to the conclusion that the RADAR sensor is a reliable sensor in series autonomous driving systems, both in a multi sensor framework and as a single component for localization.
Enhancing data-driven algorithms for human pose estimation and action recognition through simulation
(2020)
Recognizing human actions, reliably inferring their meaning and being able to potentially exchange mutual social information are core challenges for autonomous systems when they directly share the same space with humans. Intelligent transport systems in particular face this challenge, as interactions with people are often required. The development and testing of technical perception solutions is done mostly on standard vision benchmark datasets for which manual labelling of sensory ground truth has been a tedious but necessary task. Furthermore, rarely occurring human activities are underrepresented in these datasets, leading to algorithms not recognizing such activities. For this purpose, we introduce a modular simulation framework, which offers to train and validate algorithms on various human-centred scenarios. We describe the usage of simulation data to train a state-of-the-art human pose estimation algorithm to recognize unusual human activities in urban areas. Since the recognition of human actions can be an important component of intelligent transport systems, we investigated how simulations can be applied for his purpose. Laboratory experiments show that we can train a recurrent neural network with only simulated data based on motion capture data and 3D avatars, which achieves an almost perfect performance in the classification of those human actions on real data.
Due to decreased mobility or families living apart, older adults are especially vulnerable to the issue of social isolation. Literature suggests that technology can help to prevent this isolation. The present work addresses an approach to participate in society by sharing knowledge that is cherished. We propose the cooking recipe exchange application PrecRec for older adults to make them feel precious and valued. PrecRec has been developed and evaluated in an iterative process with eleven older adults. The results show that a broad perspective has to be taken into account when designing such systems.
Hypermedia as the Engine of Application State (HATEOAS) is one of the core constraints of REST. It refers to the concept of embedding hyperlinks into the response of a queried or manipulated resource to show a client possible follow-up actions and transitions to related resources. Thus, this concept aims to provide a client with a navigational support when interacting with a Web-based application. Although HATEOAS should be implemented by any Web-based API claiming to be RESTful, API providers tend to offer service descriptions in place of embedding hyperlinks into responses. Instead of relying on a navigational support, a client developer has to read the service description and has to identify resources and their URIs that are relevant for the interaction with the API. In this paper, we introduce an approach that aims to identify transitions between resources of a Web-based API by systematically analyzing the service description only. We devise an algorithm that automatically derives a URI Model from the service description and then analyzes the payload schemas to identify feasible values for the substitution of path parameters in URI Templates. We implement this approach as a proxy application, which injects hyperlinks representing transitions into the response payload of a queried or manipulated resource. The result is a HATEOAS-like navigational support through an API. Our first prototype operates on service descriptions in the OpenAPI format. We evaluate our approach using ten real-world APIs from different domains. Furthermore, we discuss the results as well as the observations captured in these tests.
3D assisted 2D face recognition involves the process of reconstructing 3D faces from 2D images and solving the problem of face recognition in 3D. To facilitate the use of deep neural networks, a 3D face, normally represented as a 3D mesh of vertices and its corresponding surface texture, is remapped to image-like square isomaps by a conformal mapping. Based on previous work, we assume that face recognition benefits more from texture. In this work, we focus on the surface texture and its discriminatory information content for recognition purposes. Our approach is to prepare a 3D mesh, the corresponding surface texture and the original 2D image as triple input for the recognition network, to show that 3D data is useful for face recognition. Texture enhancement methods to control the texture fusion process are introduced and we adapt data augmentation methods. Our results show that texture-map-based face recognition can not only compete with state-of-the-art systems under the same precon ditions but also outperforms standard 2D methods from recent years.
Das ZD.BB - Digitaler Hub für kleine und mittelständische Unternehmen in der Region Stuttgart
(2020)
Die Digitale Transformation ist eines der meistdiskutierten Themen in der heutigen Geschäftswelt. Viele Unternehmen, vor allem kleine und mittelständische Unternehmen (KMU), tun sich schwer die Chancen und Risiken der Digitalisierung einzuschätzen. Mit all den Möglichkeiten und Chancen, welche die Digitalisierung birgt, droht Unternehmen, die sich vor den Entwicklungen verschließen, der Verlust ihrer Markt- und Wettbewerbsposition. Mit dem im Februar 2019 eröffneten Digital Hub ZD.BB (Zentrum Digitalisierung) besteht in der Region Stuttgart eine neue, zentrale Anlaufstelle für Fragen rund um das Thema Digitalisierung. Am ZD.BB erhalten kleine und mittelständische Unternehmen (KMU) sowie Startups für ihre digitalen Transformationsprozesse eine kompetente Beratung und Betreuung. Sie geht von der Sensibilisierung über die Analyse bis zur Lösungsentwicklung für digitale Prozesse. Mithilfe einer digitalen Qualifizierungsoffensive und mittelstandsgerechten Methoden zur Geschäftsmodellentwicklung werden Unternehmen im ZD.BB umfassend bei ihren Digitalisierungsvorhaben unterstützt. Dazu werden in Innovationslaboren, in Coworking Spaces und bei Events unterschiedliche Kompetenzen, Disziplinen, Ideen, Technologien und Kreativität vernetzt und auf diese Weise digitale Innovationen hervorgebracht.
Development work within an experimental environment, in which certain properties are investigated and optimized, requires many test runs and is therefore often associated with long execution times, costs and risks. This can affect product, material and technology development in industry and research. New digital driver technologies offer the possibility to automate complex manual work steps in a cost-effective way, to increase the relevance of the results and to accelerate the processes many times over. In this context, this article presents a low-cost, modular and open-source machine vision system for test execution and evaluates it on the basis of a real industrial application. For this purpose a methodology for the automated execution of the load intervals, the process documentation and for the evaluation of the generated data by means of machine learning to classify wear levels. The software and the mechanical structure are designed to be adaptable to different conditions, components and for a variety of tasks in industry and research. The mechanical structure is required for tracking the test object and represents a motion platform with independent positioning by machine vision operators or machine learning. An evaluation of the state of the test object is performed by the transfer learning after the initial documentation run. The manual procedure for classifying the visually recorded data on the state of the test object is described for the training material. This leads to an increased resource efficiency on the material as well as on the personnel side since on the one hand the significance of the tests performed is increased by the continuous documentation and on the other hand the responsible experts can be assigned time efficiently. The presence and know-how of the experts are therefore only required for defined and decisive events during the execution of the experiments. Furthermore, the generated data are suitable for later use as an additional source of data for predictive maintenance of the developed object.
Additive Manufacturing is increasingly used in the industrial sector as a result of continuous development. In the Production Planning and Control (PPC) system, AM enables an agile response in the area of detailed and process planning, especially for a large number of plants. For this purpose, a concept for a PPC system for AM is presented, which takes into account the requirements for integration into the operational enterprise software system. The technical applicability will be demonstrated by individual implemented sections. The presented solution approach promises a more efficient utilization of the plants and a more elastic use.
Context: Nowadays, companies are challenged by increasing market dynamics, rapid changes and disruptive participants entering the market. To survive in such an environment, companies must be able to quickly discover product ideas that meet the needs of both customers and the company and deliver these products to customers. Dual-track agile is a new type of agile development that combines product discovery and delivery activities in parallel, iterative, and cyclical ways. At present, many companies have difficulties in finding and establishing suitable approaches for implementing dual-track agile in their business context.
Objective: In order to gain a better understanding of how product discovery and product delivery can interact with each other and how this interaction can be implemented in practice, this paper aims to identify suitable approaches to dual-track agile.
Method: We conducted a grey literature review (GLR) according to the guidelines to Garousi et al.
Results: Several approaches that support the integration of product discovery with product delivery were identified. This paper presents a selection of these approaches, i.e., the Discovery-Delivery Cycle model, Now-Next-Later Product Roadmaps, Lean Sprints, Product Kata, and Dual-Track Scrum. The approaches differ in their granularity but are similar in their underlying rationales. All approaches aim to ensure that only validated ideas turn into products and thus promise to lead to products that are better received by their users.
It is essential for the success of a company to set a strategic direction in which a product offering will be developed over time to achieve the company vision. For this reason, roadmaps are used in practice. in general, roadmaps can be expressed in various forms such as technology roadmaps, product roadmaps or industry roadmaps. From the point of view of industry, the basic purpose of a roadmap is to explore, visualize and communicate the dynamic linkage between markets, products and technology.
Nowadays companies are facing increasing market dynamics, rapidly evolving technologies and shifting user expectations. Together with the adoption of lean and agile practices this situation makes it increasingly difficult to plan and predict upfront which products, services or features should be developed in the future. Consequently, many organizations are struggling with their ability to provide reliable and stable product roadmaps by applying traditional approaches. This paper aims at identifying and getting a better understanding of which measures companies have taken to transform their current product roadmapping practices to the requirements of a dynamic and uncertain market environment. This also includes challenges and success factors within this transformation process as well as measures that companies have planned for the future. We conducted 18 semi-structured expert interviews with practitioners of different companies and performed a thematic data analysis. The study shows that the participating companies are aware that the transformation of traditional product roadmapping practices to fulfill the requirements of a dynamic and uncertain market environment is necessary. The most important measures that the participating companies have taken are 1) adequate item planning concerning the timeline, 2) the replacement of a fixed time-based chart by a more flexible structure, 3) the use of outcomes to determine the items (such as features) on the a roadmap, 4) the creation of a central roadmap which allows deriving different representation for each stakeholder and department.
Context: A product roadmap is an important tool in product development. It sets the strategic direction in which the product is to be developed to achieve the company’s vision. However, for product roadmaps to be successful, it is essential that all stakeholders agree with the company’s vision and objectives and are aligned and committed to a common product plan.
Objective: In order to gain a better understanding of product roadmap alignment, this paper aims at identifying measures, activities and techniques in order to align the different stakeholders around the product roadmap.
Method: We conducted a grey literature review according the guidelines to Garousi et al.
Results: Several approaches to gain alignment were identified such as defining and communicating clear objectives based on the product vision, conducting cross-functional workshops, shuttle diplomacy, and mission briefing. In addition, our review identified the “Behavioural Change Stairway Model” that suggests five steps to gain alignment by building empathy and a trustful relationship.
