Feb 15, 2016
We just released software (executables and sources) for our accelerated Kernel Subclass Discriminant Analysis (AKSDA) method and its combination with SVM classifiers. This is an efficient dimensionality reduction and classification method, for very high-dimensional data.
AKSDA is a new GPU-accelerated, state-of-the-art C++ library (also provided as command-line executable) for supervised dimensionality reduction and classification, using multiple kernels. It greatly reduces the dimensionality of the input data, while at the same time it increases their linear separability.
Feb 10, 2016
A joint paper by INRIA, UBX and CERTH on "Semantic Event Fusion of Different Visual Modality Concepts for Activity Recognition" has been accepted for publication in the special issue on Multimodal Human Pose Recovery and Behavior Analysis of IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI).
With an impact factor of 5.781, TPAMI is one of the most prestigious journals in the areas of computer vision and image understanding, pattern analysis and recognition, with a particular emphasis on machine learning for pattern analysis.
Jan 26, 2016
The 1st International Workshop on Multimedia Analysis and Retrieval for Multimodal Interaction (MARMI2016) is organized in conjunction with the ACM Conference on Multimedia Retrieval (ICMR) 2016, New York, USA, June 6-9, 2016. IMPORTANT DATES - Submission deadline: March 1, 2016 - Notifications: March 26, 2016 - Camera ready version: April 15, 2016 - Workshop: June 6, 2016
Jan 25, 2016
The hackAIR project was launched on January 2016 in Thessaloniki by a consortium of six partners of five European countries. hackAIR will develop an open technology toolkit for citizens observatories on air quality. This toolkit aims to complement official air quality data with a number of community-driven data sources, including an easy-to-build open hardware sensor module that transmits regular air quality measurements via Bluetooth, air quality information derived from mobile phone pictures of the sky and social media, as well as a low-tech measuring setup involving vacuum cleaners and coffee filters.
Oct 19, 2015
The Dem@Care Context Descriptor ontology has been integrated in the Linked Open Vocabularies (LOV) dataset, enabling its sharing and reuse by other datasets in the Linked Data Cloud (a human-readable description of the vocabulary is available here). The ontology has been developed in the framework of the Dem@Care project and provides the vocabulary to annotate complex (high-level) activity classes with low-level observations for complex activity recognition.
For more details, please refer to the relevant paper: G.
Oct 13, 2015
The European Summer School in Information Retrieval (ESSIR) is a scientific event founded in 1990, which has given rise to a series of Summer Schools held on a regular basis to provide high quality teaching of Information Retrieval (IR) and advanced IR topics to an audience of researchers and research students. ESSIR is typically a week-long event consisting of guest lectures and seminars from invited lecturers who are recognized experts in the field.
Sep 23, 2015
CERTH-ITI participated to the interactive Surveillance Event Detection (SED) task of TRECVID 2015 with the video retrieval engine VERGE. VERGE results achieved the second place, among the submissions made by all institutions participating to TRECVID 2015 SED task (runs from 4 different institutions searching for 7 events: cell to ear, embrace, object put, people meet, people split up, poining) with interactive systems. The latest version of VERGE integrates advanced retrieval functionalities including visual similarity search, object-based retrieval and visual concept-based search in a user friendly interface.
Sep 21, 2015
MKLab had a very successful participation in the 2015 edition of the MediaEval benchmarking activity that took place in Wurzen, Germany on 14-15 September. The team participated in four tasks: Placing, Diverse Social Image Retrieval, Verifying Multimedia Use, and Synchronization of Multi-User Event Media, with the last two tasks being co-organized by members of the lab. The team achieved impressive results in two of the tasks, namely the Placing Task, in which it achieved the best performance, being the only participation to outperform the system provided by the organizers, and in the Diverse Social Image Retrieval task, where the team achieved the best performance in the single-topic queries, and the second best performance overall.
Jul 22, 2015
MKLab is happy to announce that we received support from the Hardware Grant Program of the NVIDIA Corporation to continue our research (Zampoglou et al., 2015) on the field of multimedia forensics, which is currently carried out in the context of the REVEAL EC co-funded project. In particular, we are going to explore the potential of employing Deep Learning approaches on the problem of image splicing detection. M. Zampoglou, S.
Jul 13, 2015
The H2020 project MAMEM, started in May 1st, 2015, aims at integrating people with disabilities back into society by endowing them with the critical skill of managing and authoring multimedia content using novel and more natural interface channels. These channels will be controlled by eye-movements and mental commands, significantly increasing the potential for communication and exchange in leisure and non-leisure context. MAMEM's website adopts a rather unconventional design with the intention to become the project's "