This proposal addresses recognition of object guided human actions in video data. Automation of the analysis requires advances in human motion recognition, object recognition in video as well as incorporation of action-object rela...
ver más
¿Tienes un proyecto y buscas un partner? Gracias a nuestro motor inteligente podemos recomendarte los mejores socios y ponerte en contacto con ellos. Te lo explicamos en este video
Fecha límite de participación
Sin fecha límite de participación.
Descripción del proyecto
This proposal addresses recognition of object guided human actions in video data. Automation of the analysis requires advances in human motion recognition, object recognition in video as well as incorporation of action-object relations. Recognition in video is challenging due to dramatic changes in appearance between frames, low resolution and blur. In this work I intend to incorporate information from multiple frames at all levels of processing to overcome these difficulties. Information from multiple frames will be incorporated to detect consistent object and body parts, to obtain high-detail representations of objects and actions, as well as to infer their contextual relations. This multi view approach draws its motivation from traditional super-resolution algorithms that incorporate multiple frames into a single high-resolution view of an object or person. My goal, however, is to integrate information from multiple views to obtain ``high resolution models’' to represent objects, actions, and their relationships.