Skip to main content

Department Seminars

DCS Seminar: Dr. Matteo Naccari, BBC – Research & Development

Thursday, 24th Novermber 2016 at 4pm

Venue: CS101, Computer Science

Title: Enabling the distribution of UHDTV services using the Turing codec

Abstract: The Ultra High Definition (UHD) format will provide viewers with an enhanced quality of experience. UHD is not just about more pixels but better pixels with higher spatial and temporal resolutions and pixel dynamic range. The deployment of services using UHD content (UHD Television, UHDTV services) poses several technological challenges, most notably how to guarantee the coexistence with previous services (e.g. SDTV and HDTV) and how to handle the increased volume of data associated with the UHD format. This talk will present an overview of the current status of the technology behind the deployment of UHDTV services with particular emphasis on the Turing codec which is an open source software encoder compliant with the state-of-the-art H.265/High Efficiency Video Coding (HEVC) standard. The talk will start with an overview on the current broadcasting technology for high frame rate and high dynamic range imaging. Then the focus will move on the Turing codec, its main features, encoding optimisations and use to compress UHD material. A comparison of the performance of the Turing codec with another practical implementation of the HEVC standard will be also discussed. Finally an overview of the open source project behind the Turing codec will be also provided.

DCS Seminar: Dr. Alexandros Iosifidis, Tampere University of Technology

Thursday, 17th Novermber 2016 at 4pm

Venue: CS101, Computer Science

Title: Computational Intelligence Approaches for Digital Media Analysis and Description

Abstract: Recent advances in technological equipment, like digital cameras, smart-phones, etc., have led to an increase of the available digital media, e.g., images and videos, captured every day. Moreover, the amount of data captured for professional media production (e.g., movies, special effects, etc) has dramatically increased and diversified using multiple sensors (e.g., multi-view cameras, depth sensors, very high quality images, motion capture, etc), justifying the digital media analysis as a big data analysis problem. While this fact has increased the potential of automated digital media data analysis approaches, it also generated issues that should be appropriately addressed in order to succeed. In this talk, a short overview on recent research efforts for digital media analysis using statistical machine learning and neural networks for medium-scale and large-scale settings will be given. Their application in problems such as human face/facial expression/action recognition, object detection and recognition, salient object segmentation, image and text retrieval will be described and discussed.

WISC Seminar: Dr. Enrico Steiger, Heidelberg University

Thursday, 23rd May 2016 at 4pm

Venue: CS101, Computer Science

Title: Utilizing social media data for transport planning and traffic management

Abstract: Nowadays an increasing number of digital records relating to everyday life (blog posts, text messages, images etc.) are generated by individual users. The wide availability of location aware sensor technologies facilitates the creation of these new digital geographic footprints, which represents a unique opportunity to gain a better understanding of people and their social activities reflected in social media for the study of human mobility. Within the talk novel concepts, techniques and analysis methods for the exploration of human social activities from user-generated social media data are presented. The focus is to investigate the possibilities of characterizing spatial structures and underlying human mobility patterns, in order to assess the reliability of social media information and the given spatial, temporal and semantic characteristics. The overall aim of the talk is to demonstrate the potential of how information from social media can add further geographic information and can be utilized as a proxy indicator for the inference of real world geographic phenomena in order to provide further insights into complex human mobility processes.

WISC Seminar: Prof. Marc Scott, New York University, USA

Thursday, 30th May 2016 at 4pm

Venue: CS101, Computer Science

Title: The Impact of Food Environment on NYC Public School Students: A Quasi-Experiment and Sensitivity Analysis

Abstract: In the US, childhood obesity has reached epidemic proportions, with approximately a third of the current population identified as overweight or obese. There are of course many increased health risks associated with this condition (e.g., diabetes, CVD). Students in New York City schools are no exception, following the overall populaton trend. Policies to combat obesity in children include integrating educational materials into classroom instruction and improving the food served in cafeterias within schools (some students eat two meals per day at school). However, students spend substantial time outside of the school environment, and it is unclear how the “food environment” near the student’s home and proximate to the school are related to childhood obesity. In this study, we make use of multiple datasets that provide a comprehensive inventory of food establishments in New York City and link these to student residences and schools. Building a map of the food environment, we document the relationship between this environment and a normalized student Body Mass Index (BMI) over the academic years 2009-2013. Initial multilevel models partition the total variation into student, school and census tract (neighborhood) components, explaining a small percentage of these via student demographics and the school and home food environment. The effects associated with Fast Food and Bodega (“Shops”) Establishments appear to warrant further investigation. Making use of a natural quasi-experiment, in which the majority of students change schools in the transition to middle and high school, we evaluate whether changes in this environment are associated with changes in BMI, net of other factors. A student—level change (first difference) model apportions the covariate effects into between- and within-school components (a so-called “hybrid” model), and the remaining unexplained variance is captured via random effects. The conditions needed for a plausible causal interpretation of effects are discussed. Given the size of the school population and the costs associated with new initiatives, the magnitude of the effects associated with changes in the food environment near schools are subjected to a sensitivity analysis using software developed by the authors as part of a related methodological research effort. This is joint work with researchers associated with NYU’s PRIISM Applied Statistics Center, NYU’s Institute for Education and Social Policy, and the NYU Medical Center’s Section on Health Choice, Policy and Evaluation.