- VSD: annotated data for violent scene detection, accompanied by additional violence-related audio and video annotated features (guns, explosions, blood, etc.)
- Interestingness dataset: shots and keyframes extracted from movie trailers, with annotation of interestingness levels and additional extracted low-level features
(see complete list on GoogleScholar)
- Predicting Media Interestingness Task (2016) in the MediaEval benchmarking initiative
- Affect Task on Violent Scene Detection (2011-2015) in the MediaEval benchmarking initiative
- Quaero: program promoting research and industrial innovation on technologies for automatic analysis and classification of multimedia and multilingual documents
- Machine Learning and classification, Deep Learning,
- Multimodal analysis, structuring, indexing,
- Perceptual Understanding of content
- Image processing (video indexing, image segmentation),
- Computer vision,
- Mathematical morphology.
Claire-Helene Demarty graduated from Telecom ParisTech in 1994 and received a Ph.D. degree in Computer Science, Mathematical Morphology, from Mines ParisTech in 2000. She joined the Technicolor Research & Innovation Center in 2004 as a senior researcher and became senior scientist in 2010. Located in Rennes, France, she is working on multimedia indexing technologies and perceptual understanding of content. Prior to Technicolor, she worked at LTUTechnologies (2000 – 2002) and at INRIA Rennes – IRISA (2003 – 2004), a French public research center, as a researcher in image analysis and video indexing technologies. She is author or co-author of more than 30 papers and is holding several patents. Since 2011, she has been organizing the Affect Task - Violent Scenes Detection and in 2016, she launched the Predicting Media Interestingness Task, in the MediaEval benchmark.