Databases for Behaviour Understanding, Analysis & Generation
Databases for Behaviour Understanding, Analysis & Generation
Embedded Files
A large-scale in-the-wild A/V database consisting of 298 videos (downloaded from YouTube) of around 1,225,000 frames being annotated in terms of continuous valence and arousal emotion descriptors (dimensional affect).
The largest in-the-wild A/V database consisting of 594 videos (downloaded from YouTube) of around 3,000,000 frames. It is the only existing database annotated in terms of: i) continuous valence and arousal emotion descriptors (dimensional affect); ii) seven basic expressions (categorical affect); and iii) action units.
A large database that contains 4D videos of 4,700 subjects that display a multitude of facial behaviours, with over 280,000 3D meshes in total. A big portion of these meshes have been manually annotated with 3D facial landmarks and have been categorized in the corresponding expressions.
The largest and in-the-wild A/V database consisting of 400 videos annotated in terms of 13 compound expressions, valence-arousal emotion descriptors, action units, speech, facial landmarks and attributes.
FAIR-AffectBench
FAIR-AffectBench
We propose a unified protocol for database partitioning ensuring fairness & comparability. For 6 widely used affective databases, we provide detailed demographic annotations (in terms of race, gender and age), evaluation metrics and a common framework for expression recognition, action unit detection and valence-arousal estimation. We also provide trained models with the new protocol and introduce new leaderboards.
The largest database comprising 500+ videos and 2.7+ million frames, annotated at the frame level for violence detection. DVD captures a broad range of real-world violent and non-violent events across diverse environments, lighting conditions, camera types, participant characteristics and social interactions. DVD provides fine-grained temporal annotations, enabling the identification and localisation of violent events within longer videos.
Databases for Medical Imaging and Healthcare
Databases for Medical Imaging and Healthcare
A large-scale dataset consisting of Paraclinical data (MRI and Dat Scan images), Clinical data and Epidemiological and treatment data for Parkinson's Disease prediction; the dataset consists of 925 Dat Scans and 41,528 MRIs of 78 subjects in total (55 patients and 23 non-patients).
The largest database annotated for COVID-19 (detection and severity) that consists of around 6,600 3-D CT scans (1,550 COVID-19 cases and 5,050 non-COVID-19 ones) of around 2,500,000 CT slices of more than 1,150 patients and 2,600 non-patients.
Page updated
Google Sites
Report abuse