The project "Vision for Augmented Experiences", funded by the "Fondazione Cassa di Risparmio Modena", aims to tackle some open research challenges about cultural experiences in smart cities, for the entire lifecycle of visit capture, from the video acquisition, to its processing, summarization, personalization and social sharing.
The research focuses on some new computer science techniques, still largely unexplored.
- Egocentric-vision (or ego-vision): new computer vision techniques to automatically recognize what is taken in video captured from a first person perspective. This new research filed is very hard to tackle as egocentric captured videos consist of long streams of data with a ceaseless jumping appearance, frequent changes of observer's focus and lack of hard cuts between scenes, thus requiring new methodologies for automatic analysis and understanding. It s' important to recognize in real time what the user sees and and interact with him to provide information and knowledge and create a personalized experience. New models will be used for recognition of gestures and social behavior of people.
- Semantic annotation of content and knowledge extraction from captured video, to achieve recognition of objects, concepts and visual events through the extraction of low level descriptors as visual features, use of statistical and pattern recognition algorithms in machine learning classification with noisy training data automatically extracted from the web, and the use of audio, visual and textual cross-media descriptors, through artificial intelligence methods to provide knowledge management, such as data mining and topic extraction.
- Augmented and customized video rendering, to create user experience personalized video summaries: in particular video summarization techinques are used to mantain the relevant semantic content from hours of original video, summarizing it in a few minutes, and video personalization techinques aim to enrich the video with additional heterogeneous media data, on the basis of user profile. Recommendation and profiling methods will be integrated with automatic content-based and collaborative-based solutions .
The application of these techniques in the fields of Cultural Heritage is extremely innovative: the idea is to take a cue from social marketing techniques to process and enrich cultural experiences video captured through machine learning advanced solutions in order to achieve effective deepening adaptive solutions . Web and mobile applications are implemented using human-machine interaction technologies to enrich the traditional web and mobile techinques.
A knowledge management platform will be implemented for the storage and exchange of information between the different modules of the project. For this purpose, new solutions and approaches for the management of Big Data will be developed to automatically connect metadata and visual content to the web knowledge repositories.