Marc Ritter

Produce. annotate. archive. repurpose --: accelerating the composition and metadata accumulation of tv content

Supporting most aspects of a media providers real workflows such as production, distribution, content description, archiving, and re-use of video items, we developed a holistic framework to solve issues such as lack of human resources, necessity of parallel media distribution, and retrieving previously archived content through editors or consumers.

An Extensible Tool for the Annotation of Videos Using Segmentation and Tracking

Due to massive amount of data, the description of audiovisual media by metadata nowadays can benefit by the support of (semi-)automatic methods during the annotation process. The presented tool enables the user to mark, interactively segment and track preselected objects. An integrated shot detection splits the video into disjoint parts, for instance to circumvent the semi-automated tracking of objects across shot boundaries. Arbitrary application dependent custom image processing chains can be created in conjunction with the research framework AMOPA. Created data is exported in compliance to MPEG7-DAVP.

Annotate. Train. Evaluate. A Unified Tool for the Analysis and Visualization of Workflows in Machine Learning Applied to Object Detection

The development of classifiers for object detection in images is a complex task that comprises the creation of representative and potentially large datasets from a target object by repetitive and time-consuming intellectual annotations, followed by a sequence of methods to train, evaluate and optimize the generated classifier. This is conventionally achieved by the usage and combination of many different tools. Here, we present a holistic approach to this scenario by providing a unified tool that covers the single development stages in one solution to facilitate the development process. We prove this concept by the example of creating a face detection classifier.

Improving Transcript-Based Video Retrieval Using Unsupervised Language Model Adaptation

One challenge in automated speech recognition is to determine domain-specific vocabulary like names, brands, technical terms etc. by using generic language models. Especially in broadcast news new names occur frequently. We present an unsupervised method for a language model adaptation, which is used in automated speech recognition with a two-pass decoding strategy to improve spoken document retrieval on broadcast news. After keywords are extracted from each utterance, a web resource is queried to collect utterance-specific adaptation data. This data is used to augment the phonetic dictionary and adapt the basic language model. We evaluated this strategy on a data set of summarized German broadcast news using a basic retrieval setup.

A support framework for automated video and multimedia workflows for production and archive

The management of the massive amount of data in video- and multimedia workflows is a hard and expensive work that requires much personnel and technical resources. Our flexible and scalable open source middleware framework offers solution approaches for the automated handling of the ingest and the workflow by an automated acquisition of all available information. By using an XML format to describe the processes, we provide an easy, fast and well-priced solution without the need for specific human skills.

OphthalVis: making data analytics of optical coherence tomography reproducible

In this paper, we discuss the issues of the current state of the art in optical coherence tomography with respect to reproducibility. We present our findings about the internal computations and data storage methods of the currently used devices. The gained knowledge was used to implement a tool to read a variety of OCT file formats and reproduce the visualizations used in daily clinical routine.

Comparing Visual Data Fusion Techniques Using FIR and Visible Light Sensors to Improve Pedestrian Detection

Pedestrian detection is an important field in computer vision with applications in surveillance, robotics and driver assistance systems. The quality of such systems can be improved by the simultaneous use of different sensors. This paper proposes three different fusion techniques to combine the advantages of two vision sensors -- a far-infrared (FIR) and a visible light camera. Different fusion methods taken from various levels of information representation are briefly described and finally compared regarding the results of the pedestrian classification.

Adaptive Sensor Data Fusion for Efficient Climate Control Systems

Thousands of data centres are using traditional air-conditioned cooling concepts for the entire payload. Most of these data centres include multiple hardware generations and different types of IT-infrastructure components, i.e. storage, compute, and network devices. In the context of Green-IT, an efficient and safe parameterization of the air-conditioning-system is essential - keep the temperature as low as necessary, but not too low. Usually, only a few amount of temperature sensors are available to handle these important control cycles. But in order to optimise the cooling capacity, several scenario-specific parameters have to be considered, including the shape of the room, air flow, or component placements. In this context, the TU Chemnitz develops novel concepts to improve this process. We are using local sensor capabilities within the hardware components and combine these information with actual system loads to create an extended knowledge base, which also provides adaptive learning features. First measurement scenarios show huge optimisation potential. The respective trade-off between power consumption and cooling capacity results in significant cost savings.

Rapid Model-Driven Annotation and Evaluation for Object Detection in Videos

Nowadays, the annotation of ground truth and the automated localisation and validation of objects in audiovisual media plays an essential role to keep pace with the large data growth. A common approach to train such classifiers is to integrate methods from machine learning that often demand multiple thousands or millions of samples. Therefore, we propose two components. The first constraints the annotation space by predefined models and allows the creation of ground truth data while providing opportunities to annotate and interpolate objects in keyframes or in-between by granting a user-friendly frame-wise access. The graphical user-interface of the second component focuses on the rapid validation of automatically pre-classified object instances in order to alter the assignment of the class label or to remove false-positives to clean-up the result list which has been successfully applied on the task of Instance Search within the TRECVid evaluation campaign.

Cognitive Tools for Design Engineers: A Framework for the Development of Intelligent CAD Systems

Abstract Great design engineers are highly creative and unorthodox individuals who invent novel solutions that satisfy a set of constraints that are often ill-defined and customer driven. Designers use many tools to develop their designs, such as computer aided design (CAD) systems, that do not support the cognition that drives the design process. This paper develops the cognitive psychological background, a state of the practice based rationale for CAD enhancement and the research framework for cognitive CAD tools that support the design engineer during the creative problem solving process through reasoning and meaningful design alternatives. The research framework presented here was initially created for the development of cognitive tools for mechanical design but is transferable to other design disciplines. At the core of the research plan are the development and implementation of an artificial memory that is interpreted with real-time data analyses supported by machine learning, and made accessible to the design engineer through interaction design for intelligent CAD (iCAD).