Refine
Year of publication
- 2018 (88) (remove)
Document Type
- Conference Proceeding (88) (remove)
Keywords
Institute
- Fakultät Bauingenieurwesen (1)
- Fakultät Informatik (13)
- Fakultät Maschinenbau (3)
- Fakultät Wirtschafts-, Kultur- und Rechtswissenschaften (2)
- Institut für Angewandte Forschung - IAF (3)
- Institut für Optische Systeme - IOS (9)
- Institut für Strategische Innovation und Technologiemanagement - IST (5)
- Institut für Systemdynamik - ISD (6)
- Institut für professionelles Schreiben - IPS (2)
- Konstanz Institut für Corporate Governance - KICG (2)
Visualization-Assisted Development of Deep Learning Models in Offline Handwriting Recognition
(2018)
Deep learning is a field of machine learning that has been the focus of active research and successful applications in recent years. Offline handwriting recognition is one of the research fields and applications were deep neural networks have shown high accuracy. Deep learning models and their training pipeline show a large amount of hyper-parameters in their data selection, transformation, network topology and training process that are sometimes interdependent. This increases the overall difficulty and time necessary for building and training a model for a specific data set and task at hand. This work proposes a novel visualization-assisted workflow that guides the model developer through the hyper-parameter search in order to identify relevant parameters and modify them in a meaningful way. This decreases the overall time necessary for building and training a model. The contributions of this work are a workflow for hyper-parameter search in offline handwriting recognition and a heat map based visualization technique for deep neural networks in multi-line offline handwriting recognition. This work applies to offline handwriting recognition, but the general workflow can possibly be adapted to other tasks as well.
Algorithms for calculating the string edit distance are used in e.g. information retrieval and document analysis systems or for evaluation of text recognizers. Text recognition based on CTC-trained LSTM networks includes a decoding step to produce a string, possibly using a language model, and evaluation using the string edit distance. The decoded string can further be used as a query for database search, e.g. in document retrieval. We propose to closely integrate dictionary search with text recognition to train both combined in a continuous fashion. This work shows that LSTM networks are capable of calculating the string edit distance while allowing for an exchangeable dictionary to separate learned algorithm from data. This could be a step towards integrating text recognition and dictionary search in one deep network.
Offline handwriting recognition systems often use LSTM networks, trained with line- or word-images. Multi-line text makes it necessary to use segmentation to explicitly obtain these images. Skewed, curved, overlapping, incorrectly written text, or noise can lead to errors during segmentation of multi-line text and reduces the overall recognition capacity of the system. Last year has seen the introduction of deep learning methods capable of segmentation-free recognition of whole paragraphs. Our method uses Conditional Random Fields to represent text and align it with the network output to calculate a loss function for training. Experiments are promising and show that the technique is capable of training a LSTM multi-line text recognition system.
Optical surface inspection: A novelty detection approach based on CNN-encoded texture features
(2018)
In inspection systems for textured surfaces, a reference texture is typically known before novel examples are inspected. Mostly, the reference is only available in a digital format. As a consequence, there is no dataset of defective examples available that could be used to train a classifier. We propose a texture model approach to novelty detection. The texture model uses features encoded by a convolutional neural network (CNN) trained on natural image data. The CNN activations represent the specific characteristics of the digital reference texture which are learned by a one-class classifier. We evaluate our novelty detector in a digital print inspection scenario. The inspection unit is based on a camera array and a flashing light illumination which allows for inline capturing of multichannel images at a high rate. In order to compare our results to manual inspection, we integrated our inspection unit into an industrial single-pass printing system.
Digitale Signaturen zum Überprüfen der Integrität von Daten, beispielsweise von Software-Updates, gewinnen zunehmend an Bedeutung. Im Bereich der eingebetteten Systeme kommen derzeit wegen der geringen Komplexität noch überwiegend symmetri-sche Verschlüsselungsverfahren zur Berechnung eines Authentifizierungscodes zum Einsatz. Asym-metrische Kryptosysteme sind rechenaufwendiger, bieten aber mehr Sicherheit, weil der Schlüssel zur Authentifizierung nicht geheim gehalten werden muss. Asymmetrische Signaturverfahren werden typischerweise zweistufig berechnet. Der Schlüssel wird nicht direkt auf die Daten angewendet, sondern auf deren Hash-Wert, der mit Hilfe einer Hash-funktion zuvor berechnet wurde. Zum Einsatz dieser Verfahren in eingebetteten Systemen ist es erforder-lich, dass die Hashfunktion einen hinreichend gro-ßen Datendurchsatz ermöglicht. In diesem Beitrag wird eine effiziente Hardware-Implementierung der SHA-256 Hashfunktion vorgestellt.
The Role of Support-Activities for the successful Implementation of Internal Corporate Accelerators
(2018)
Today’s markets are characterized by fast and radical changes, posing an essential challenge to established companies. Startups, yet, seem to be more capable in developing radical innovations to succeed in those volatile markets. Thus, established companies started to experiment with various approaches to implement startup-like structures in their organization. Internal corporate accelerators (ICAs) are a novel form of corporate venturing, aiming to foster bottom-up innovations through intrapreneurship. However, ICAs still lack empirical investigations. This work contributes to a deeper understanding of the interface between the ICA and the core organization and the respective support activities (resource access and support services) that create an innovation-supportive work environment for the intrapreneurial team. The results of this qualitative study, comprising 12 interviews with ICA teams out of two German high-tech companies, show that the resources provided by ICAs differ from the support activities of external accelerators. Further, the study shows that some resources show both supportive as well as obstructive potential for the intrapreneurial teams within the ICA.
This work proposes a construction for low-density parity-check (LDPC) codes over finite Gaussian integer fields. Furthermore, a new channel model for codes over Gaussian integers is introduced and its channel capacity is derived. This channel can be considered as a first order approximation of the additive white Gaussian noise channel with hard decision detection where only errors to nearest neighbors in the signal constellation are considered. For this channel, the proposed LDPC codes can be decoded with a simple non-probabilistic iterative decoding algorithm similar to Gallager's decoding algorithm A.
Generalized concatenated (GC) codes with soft-input decoding were recently proposed for error correction in flash memories. This work proposes a soft-input decoder for GC codes that is based on a low-complexity bit-flipping procedure. This bit-flipping decoder uses a fixed number of test patterns and an algebraic decoder for soft-input decoding. An acceptance criterion for the final candidate codeword is proposed. Combined with error and erasure decoding of the outer Reed-Solomon codes, this bit-flipping decoder can improve the decoding performance and reduce the decoding complexity compared to the previously proposed sequential decoding. The bit-flipping decoder achieves a decoding performance similar to a maximum likelihood decoder for the inner codes.
„Nackte Tatsachen genügen nicht, um den Leser bis zum Schluss bei der Stange zu halten. [Auch der] Verfasser von [Fach- und] Sachtexten hat das Recht – wenn nicht gar die Pflicht –, die Wunder der Erzählkunst für sich zu erschließen, um seine Texte zu einer interessanten und vergnüglichen Lektüre zu machen.“ – Der Vortrag geht der Frage nach, ob diese Forderung des vielgelesenen Ratgeberautors Sol Stein berechtigt ist und inwiefern sie eingelöst werden kann. Zu diesem Zweck werden Stil- und Kompositionsmittel aus den Bereichen Lexik, Grammatik, Syntax, Stilistik, Erzählstrategie und Textorganisation anhand von Formulierungsbeispielen vorgestellt und auf ihre Funktion in Erzählliteratur oder Journalismus hin untersucht. In einem zweiten Schritt wird jeweils nach der Übertragbarkeit dieser sprachlichen Mittel auf Fachtexte gefragt. Anhand parallel ausgerichteter Beispiele wird eine mögliche verlagerte Funktion im Wissenschaftskontext skizziert. Der erhoffte Erkenntnisgewinn des Vortrags betrifft sowohl das Verfassen eigener Publikationen als auch die Vermittlung von Schreibfähigkeiten für künftige Fachautor*innen, weshalb die Formulierungsbeispiele verschiedensten, auch technischen und naturwissenschaftlichen Domänen entnommen sind.
(Literaturangabe: Stein, Sol: Über das Schreiben [1995], dt. Waltraud Götting, 10. Aufl., Zweitausendeins, Frankfurt a. M., 2006, zit. S. 49)