Logo des Repositoriums

Auflistung nach Autor:in "Kowerko, Danny"

1 - 6 von 6
  • Textdokument
    Acoustic Event Classification Using Convolutional Neural Networks
    (INFORMATIK 2017, 2017) Kahl, Stefan; Hussein, Hussein; Fabian, Etienne; Schloßhauer, Jan; Thangaraju, Enniyan; Kowerko, Danny; Eibl, Maximilian
    The classification of human-made acoustic events is important for the monitoring and recognition of human activities or critical behavior. In our experiments on acoustic event classification for the utilization in the sector of health care, we defined different acoustic events which represent critical events for elderly or people with disabilities in ambient assisted living environments or patients in hospitals. This contribution presents our work for acoustic event classification using deep learning techniques. We implemented and trained various convolutional neural networks for the extraction of deep feature vectors making use of current best practices in neural network design to establish a baseline for acoustic event classification. We convert chunks of audio signals into magnitude spectrograms and treat acoustic events as images. Our data set contains 20 different acoustic events which were collected in two different recording sessions combining human and environmental sounds. Our results demonstrate how efficient convolutional neural networks perform in the domain of acoustic event classification.
  • Konferenzbeitrag
    Aufbereitung augenmedizinischer Bild-, Patienten- und Diagnosedaten zum Zwecke der Forschung - Ethikrichtlinen und deren praktische Umsetzung
    (Mensch und Computer 2017 - Workshopband, 2017) Kowerko, Danny; Rößner, Miriam; Kahl, Stefan; Herms, Robert; Eibl, Maximilian; Engelmann, Katrin
    In der vorliegenden Arbeit stellen die Authoren einen technischen Workflow vor, der darstellt wie in der Praxis gesetzliche Vorgaben im Bezug auf ethische Fragestellungen umgesetzt werden können. Dabei wird auf die rechtlichen Grundlagen auf Bundesebene eingegangen, aber auch auf die Besonderheiten auf Länderebene und der lokalen Umsetzung. Am Fallbeispiel der Kooperation zwischen der Juniorprofessur Media Computing an der TU Chemnitz und der Augenklinik des Klinikum Chemnitz gGmbH zeigen wir dabei welche Vorgaben seitens des Ethikbeauftragten einzuhalten waren hinsichtlich der Anonymisierung von Patientendaten, der Verschlüsselung, dem Transport/Transfer von der Klinik an die Universität, Speicherung und Zugriffsrechte der Daten. Eingegangen wird insbesondere auf unterschiedliche Aspekte in der retrospektiven Forschung mit Patientendaten. Damit soll insbesondere Einsteigern auf dem Gebiet der Forschung mit klinischen Daten ein erster Einblick ermöglicht werden.
  • Textdokument
    Evaluation of CNN-based algorithms for human pose analysis of persons in red carpet scenarios
    (INFORMATIK 2017, 2017) Kowerko, Danny; Richter, Daniel; Heinzig, Manuel; Kahl, Stefan; Helmert, Stefan; Brunnett, Guido
    We evaluate two CNN-based algorithms for keypoint-based human pose analysis on two image test sets containing red carpet scenarios, one taken under controlled conditions in a TV studio environment and another more heterogeneous data set taken from FlickR without any restriction but to contain a red carpet. We focus on the pose of persons standing directly on the red carpet. A web application is presented allowing collaborative work to confirm or modify already pre-localised body keypoints given from the method presented in [Ca17]. These annotations helped to quickly define ground truth for the subsequent evaluation of several hundreds of persons standing on a red carpet. An own evaluation formalism is presented that adopts to the size of the respective keypoints. The TV studio data set includes coarsely defined body and head poses. Using the angular information, we are able to quantitatively define the optimum head pose angle range and limitations of facial keypoint determination.
  • Textdokument
    Fast and accurate creation of annotated head pose image test beds as prerequisite for training neural networks
    (INFORMATIK 2017, 2017) Kowerko, Danny; Manthey, Robert; Heinz, Marcel; Kronfeld, Thomas; Brunnett, Guido
    In this paper we present an experimental setup consisting of 36 cameras on 4 height levels covering more than half space around a centrally sitting person. The synchronous image release allows to build a 3D model of the human torso in this position. Using this so-called body scanner we recorded 36 different positions giving in total 1296 images in several minutes obtaining tens to hundreds of different pitch-roll-yaw head pose combinations with very high precision of less than +-5. From annotation of 7 facial keypoints (ears, eyes, nose, corners of the mouth) in the 36 calculated 3D models of a human head/upper body, we automatically get 1296 x 7 2D facial landmark points saving a factor 36 in annotation time. The projection of the 3D model to the camera provides a foreground/background separation mask of the person in each image usable for data set augmentation e.g. by inserting different backgrounds (required for training convolutional neural networks, CNNs). Moreover, we utilize our 3D model in combination with textures to create realistic images of the pitch-roll-yaw range not assessed in experiments. This interpolation is ad hoc applicable to a subset of 10 central out of 36 total camera views where fine-grained interpolation of head poses is possible. Using interpolation and background masks for background exchange enables us to augment the data set easily by a factor of 1000 or more knowing precisely pitch, roll, yaw and the 7 annotated facial keypoints in each image.
  • Textdokument
    Preparing clinical ophthalmic data for research application
    (INFORMATIK 2017, 2017) Rößner, Miriam; Kahl, Stefan; Engelmann, Katrin; Kowerko, Danny
    This paper presents an analysis of clinical examination, diagnostic and patient data belonging to persons with eye diseases like age-related macular degeneration (AMD). Our purpose is to investigate potential correlations of extracted features to discover their impacts on the disease. This is a first step to the predictability of the progression of AMD based on a heterogeneous data set. We focus on the visual acuity as reasonable indicator for the progression of this disease and analyse its temporal trend to classify patients in winners, stabilisers and losers.We describe the retrieval of textual medical reporting data for optical coherence tomography images and evaluate the machine-readable categorisation of these texts. Additionally, we address the topic of ethical guidelines for the work with patients’ data and discuss the potential and limitations of our data set in the context of obtaining structured (mass) data for training neural networks as future perspective.
  • Textdokument
    WS34 - Deep Learning in heterogenen Datenbeständen
    (INFORMATIK 2017, 2017) Kowerko, Danny; Kahl, Stefan
    Deep learning techniques, especially artificial neural networks, have become irreplaceable in almost every aspect of modern information science. Breakthrough technologies evolve rapidly, driven by researchers with both, scientific and economic backgrounds. This workshop is a platform for students, post-docs, innovative enterprises and experts from Germany who present their latest works and demo applications. Recent advanced in the field of deep learning and their impact on research projects and economic endeavors are at the center of submitted papers and presentations. An active debate focusing on current work-in-progress, future research as well as chances and opportunities of deep learning is complemented by the discussion regarding the generation, processing and publishing of large heterogeneous datasets for research purposes. The presented contributions span a wide variety of deep learning applications – from robotics to audio and text retrieval, from human pose estimation to medical data processing. This not only demonstrates how important deep learning techniques have become for almost every area of research, it also shows the importance of scientific transparency.Without the efforts of countless researchers around the globe who published their work and complemented it with code repositories and extensive documentation, some of the presented applications could not have been implemented. This reminds us: An active deep learning community is vital for the success of innovative data processing routines and with that, forms the foundation of a steady evolution powered by scientific research. We would like to thank everyone who participated in this workshop, especially the authors and presenters who contributed to the success of this novel format.