In particular, feature extraction, image segmentation, object recognition, motion analysis and scene understanding are topics which have been advanced by the research carried out at the PeRCeiVe Lab. Applications of the devised methods are mainly in the areas of environmental video monitoring, event detection, ﬁne-grained visual categorisation and medical image analysis.
Pattern recognition research within the PeRCeiVe Lab has mainly investigated decision trees, deep learning methods, Hidden Markov Models and mixture models for applications in the medical imaging domain, bioinformatics, eye tracking and video content analysis. Methods for automatic generation of large scale annotated datasets and for reliable performance evaluation in pattern recognition are also under investigation.
The focus of the research under this topic is on devising techniques for the analysis, processing, indexing and ﬁltering of multimedia data. This includes a broad range of problems, from content extraction and understanding to similarity assessment mainly for information retrieval applications. Particular emphasis has been recently given to multimedia recorded for environment monitoring and to integration of heterogeneous multimedia data.
With the explosion of visual data on the web, computer vision researchers have started to direct their attention towards the development of non-parametric data-driven models for solving complex visual tasks (which at the moment are far from being solved) such as object detection, recognition and scene understanding and promising preliminary results have been obtained.