Abstract: Content-based analysis and retrieval of digital images found in scientific articles is often hindered by images consisting of multiple subfigures (compound figures). We address this problem by proposing a method (ComFig) to automatically classify and separate compound figures, which consists of two main steps: (i) a supervised compound figure classifier (ComFig classifier) discriminates between compound and non-compound figures using task-specific image features; and (ii) an image processing algorithm is applied to predicted compound images to perform compound figure separation (ComFig separation). The proposed ComFig classifier is shown to achieve state-of-the-art classification performance on a published dataset. Our ComFig separation algorithm shows superior separation accuracy on two different datasets compared to other known automatic approaches. Finally, we propose a method to evaluate the effectiveness of the ComFig chain combining classifier and separation algorithm, and use it to optimize the misclassification loss of the ComFig classifier for maximal effectiveness in the chain.

Abstract: We propose an image processing algorithm to automatically separate compound figures appearing in scientific articles. We classify compound images into two classes and apply different algorithms for detecting vertical and horizontal separators to each class: the edge-based algorithm aims at detecting visible edges between subfigures, whereas the band-based algorithm tries to detect whitespace separating subfigures (separator bands). The proposed algorithm has been evaluated on two datasets for compound figure separation (CFS) in the biomedical domain and compares well to semi-automatic or more comprehensive state-of-the-art approaches. Additional experiments investigate CFS effectiveness and classification accuracy of various classifier implementations.

Abstract: Fitness- or exergames are ubiquitously available, but often lack the main ingredient of successfully gamified systems: fun. This can be attributed to the typical way of designing such games -- highly focusing on specific physical activities, thus, gamifying fitness. Instead, we propose a novel alternate approach to improve motivation for exergaming, which we call fitnessification: integrating physical exercise into very popular games that have been developed keeping fun in mind and frequently are played for long periods of time -- so-called AAA games. In order to evaluate this concept, we have conducted a comparative study examining voluntary participants' reactions to testing an ergometer controlled casual game as well as a modified AAA game. Results indicate strong tendencies of players preferring the newly introduced AAA approach over the casual fitness game.

Abstract: The computer communication research community shows significant interest in the paradigm of Information-Centric Networking (ICN). Continuously, new proposals for ICN-related challenges (caching, forwarding, etc.) are published. However, due to a lack of a readily available testbed, the majority of these proposals is evaluated either by theoretical analysis and/or by conducting network simulations potentially masking further challenges that are not observable in synthetic environments. Therefore, this article presents a framework for an ICN testbed using low-budget physical hardware with little deployment and maintenance effort for the individual researcher; specifically, Named Data Networking is considered. The employed hardware and software are powerful enough for most research projects, but extremely resource intensive tasks may push both components towards their limits. The testbed framework is based on well established open source software and provides the tools to readily investigate important ICN characteristics on physical hardware emulating arbitrary network topologies. The article discusses the testbed architecture and provides first results obtained from emulations that investigate the performance of various forwarding strategies. The results indicate that further challenges have to be overcome when heading towards a real-world deployment of ICN-based communication.

Abstract: Adaptive content delivery is the state-of-the-art in real-time multimedia streaming. Leading streaming approaches, e.g., MPEG-DASH and Apple HLS, have been developed for classical IP-based networks, providing effective streaming by means of pure client-based control and adaptation. However, the research activities of the Future Internet community adopt a new course that is different from today's host-based communication model. So-called Information-Centric Networks are of considerable interest and are advertised as enablers for intelligent networks, where effective content delivery is to be provided as an inherent network feature. This paper investigates the performance gap between pure client-driven adaptation and the theoretical optimum in the promising Future Internet architecture Named Data Networking (NDN). The theoretical optimum is derived by modeling multimedia streaming in NDN as a fractional Multi-Commodity Flow Problem and by extending it taking caching into account. We investigate the multimedia streaming performance under different forwarding strategies, exposing the interplay of forwarding strategies and adaptation mechanisms. Furthermore, we examine the influence of network inherent caching on the streaming performance by varying the caching polices and the cache sizes.

Abstract: In this report we describe our approach to the fully automatic Ad-hoc video search task for TRECVID 2016. We describe how we obtain training data from the web, create according CNN models for the provided queries and use them to classify keyframes from a custom sub-shot detection method. The resulting classifications are fed into a Lucene index in order to obtain the shots that match the query. We also discuss our results and point out potentials for further improvements.

Abstract: Real-time entertainment (mainly audio/video streaming) is responsible for the largest traffic share in today's networks. Social and entertainment platforms such as YouTube, Netflix and Facebook provide a tremendous amount of multimedia content to their global customers via the Internet. With the ever growing popularity of these services, the Internet is struggling to suffice the continuously increasing requirements demanded by applications. In particular, the demands go far beyond the intent of the Internet's original design. Architectural and legacy design choices lead to issues, the solutions to which are neither efficient nor elegant. One approach to tackle these challenges is Information-Centric Networking (ICN), a new concept for today's Internet. The idea is to base the network's principal communication model on the most important item, namely the content to be transferred. This novel concept provides significant opportunities to enhance networking. In this thesis we investigate how ICN can be used as an enabler for effective multimedia dissemination. As a first step we analyse the technology's characteristic capabilities and their potential benefits for content distribution in future networks. We develop an analytical model taking account of the main building blocks (network-inherent caching, multi-path forwarding) and compare the obtained upper bound to the current state of ICN considering the scenario of pull-based adaptive multimedia streaming. The results show that there exists a significant gap between the promised and the realized performance, largely caused by ineffective Interest forwarding strategies. Therefore, we design and implement a novel probability-based forwarding strategy named Stochastic Adaptive Forwarding (SAF), which provides effective multi-path forwarding, identifies unknown cached content replicas and deals with local topology changes without guidance from the routing plane. The results indicate that SAF brings ICN one step closer towards effective content distribution. In particular, we show that it is important to consider context information in the forwarding plane. This includes content characteristics and application demands. SAF is the first strategy that takes account of context information that can be supplied by the network operators. Furthermore, this work provides a framework for a testbed that can be used by researchers to readily deploy an ICN-based testbed. This allows researchers to conduct experiments on physical hardware providing deeper insights on proposed algorithms than network simulations or analytical methods could ever do. We use the testbed to validate our results concerning multimedia delivery in ICN, and conduct network emulations investigating the performance of SAF and its competitors. Furthermore, we compare the results of network emulations to the findings obtained from simulations to assess their validity. Both simulations and emulations show that our SAF approach provides a significant step towards effective multimedia content distribution in ICN.

Abstract: Today, the global share and increase of Internet traffic is largely caused by multimedia delivery, mainly encompassing video, audio and image sharing on social, news, and entertainment platforms. This fact is well known to the Internet research community, which tries to counteract by increasing the content delivery efficiency. So-called Information-Centric Networks (ICN) are of considerable interest, advertised as enablers for intelligent networks, where effective delivery is to be provided as an inherent network feature. Most research proposals in this area are evaluated in simulated environments, using simulation frameworks such as OMNeT++ or ns-3. However, simulations always have shortcomings and cannot substitute measurements in physical networks. In this demonstration, we show how to readily set up an ICN-based testbed using low-budget single-board computers to conduct comprehensive emulations. We choose the scenario of pull-based adaptive video delivery as a showcase and evaluate the performance of different client-based adaptation mechanisms at the application level and different content forwarding strategies at the network level. All of the presented tools and visualization features are provided as open source contributions to the community.

Abstract: This report summarizes the presentations and discussions of the special session entitled “Perspectives on Multimedia Analytics” at MMM 2016, which was held in Miami, Florida on January 6, 2016. The special session consisted of four brief paper presentations, followed by a panel discussion with questions from the audience. The session was organized by Björn Þór Jónsson and Cathal Gurrin, and chaired and moderated by Klaus Schoeffmann. The goal of this report is to record the conclusions of the special session, in the hope that it may serve members of our community who are interested in Multimedia Analytics.

Abstract: Content-Centric Networking (CCN) research addresses the mismatch between the modern usage of the Internet and its outdated architecture. Importantly, CCN routers use various caching strategies to locally cache content frequently requested by end users. However, it is unclear which content shall be stored and when it should be replaced. In this work, we employ novel techniques towards intelligent administration of CCN routers. Our approach allows for autonomous switching between existing strategies in response to changing content request patterns using rule-based stream reasoning framework LARS which extends Answer Set Programming for streams. The obtained possibility for flexible router configuration at runtime allows for faster experimentation and may result in significant performance gains, as shown in our evaluation.

Abstract: We propose a novel video browsing approach that aims at optimally integrating traditional, machine-based retrieval methods with an interface design optimized for human browsing performance. Advanced video retrieval and filtering (e.g., via color and motion signatures, and visual concepts) on a desktop is combined with a storyboard-based interface design on a tablet optimized for quick, brute-force visual inspection. Both modules run independently but exchange information to significantly minimize the data for visual inspection and compensate mistakes made by the search algorithms.