Sunday, August 25th, Morning


Performance Evaluation in Document Image Analysis
Apostolos Antonacopoulos, Basilis Gatos, Stefan Pletschacher
Performance evaluation, based on objective measures and representative datasets, is crucial to making real progress in any field. In Document Image Analysis, a field with many different practical applications, there is a large number of methods proposed but it is not clear, faced with a particular type of document, which method is more applicable or how an existing method can be improved to be better suited to a given application. This tutorial of will cover the key issues in performance evaluation in the most widely researched but, at the same time, more difficult to assess areas of Document Image Analysis. All aspects of performance evaluation will be examined, from collecting a representative sample to ground truthing to defining evaluation metrics and scenarios to interpreting the results. Participants will be given copies of software tools and will be guided through example ground truthing and evaluation workflows.


Building Fast High-Performance Recognition Systems with Recurrent Neural Networks and LSTM
Thomas Breuel, Volkmar Frinken, Marcus Liwicki
In this half-day tutorial several Recurrent Neural Networks (RNNs) and their application to Pattern Recognition will be described. First, a brief history of RNNs is presented. Next, several problems of simple RNNs are described and the Long Short-Term Memory (LSTM) is presented as a solution for those problems. For a better understanding of the network, its behaviour on several toy problems and real-world PR-applications is investigated. Finally, extended architectures, such as the bi- and multi- directional LSTM will be proposed and their application to speech, handwriting and other PR-domains will be given. Existing Open-Source Toolkits implementing the LSTM and some extensions will be presented and an introduction of how to use these tools will be given.