APPIS will host two tutorials on Machine Learning on January 9th, 2018.
The participation to the tutorial is not included in the conference fee, and costs 40EUR per tutorial.

January 9th – 14:00-17:30

Deep Learning and applications to activity recognition from Egocentric Photostreams – Prof. Petia Radeva, University of Barcelona, with Marc Bolaños and Estefania Talavera

This tutorial aims to be an overview of Deep Learning techniques with a special focus on Multimodal (video, images and text) learning models, both considering their definition and applicability. First, we will introduce the basics of Deep Learning techniques, a field that is currently revolutionising the Computer Vision field. Then, we will delve deeper into the modelling of multimodal techniques, considering problems that involve the use of both images and text (e.g. image description, video description or visual question answering). Finally, we will present an overview on how the introduced techniques have been applied to activity recognition and social analysis. We will discuss a specific application in the field of Egocentric Photostreams analysis as a unique tool to analyse the person’s activities, environment and social context. We will address the following problems: temporal segmentation, activities recognition and ego-video description. Within this last part, we will focus on describing how physical and social activities are recognized from our data, and how semantic information can be extracted. We will finish with a discussion on potential applications of Deep learning and activities recognition to two health problems: augmenting memory of mild cognitive impairment patients and preventing fragility of elder people.