GIOVANNI MARIA FARINELLA

Department of Mathematics and Computer Science - University of Catania


GMF Home Page --> POSTDOCTORAL OPPORTUNITIES

Closed Position

Domain Adaptation for Egocentric Vision Systems

In accordance with the aims of the project “First Person (Egocentric) Vision for Scene Understanding”, Piano della Ricerca 2016-2018 – linea di intervento 2, this Post-Doctoral programme is aimed to the study and development of innovative algorithms to be employed for Egocentric Vision Systems that must be able to understand the observed scene from the point of view of the agent wearing the camera. In particular, the study will be focused on “Domain Adaptation” techniques based on Deep Learning to be exploited to train egocentric vision system with simulated visual data and able to generalize on real visual data acquired with mobile and wearable imaging devices.

Application Deadline: 02 July 2019

See Application Procedure Details


Closed Position

Scene Understanding and Behavioural Analysis from Egocentric Visual Data

The aim of the research project is the study, design and development of Egocentric (First-Person) Vision algorithms able to localize users in the different areas (e.g., rooms) of an environment (eg. Home, Airport, cultural/natural site) by exploiting images acquired with a wearable camera. The algorithms should be based on recent Computer Vision and Machine Learning (e.g., Deep Learning) methods. The algorithms should be able to understand the scene and to recognize the different contexts where visitors move, as well as the observed objects. The 2D position of the visitors with respect to the map of the environments, and the 3D position of the wearable cameras have to be inferred from egocentric images/video. The inferred data should be used to develop behavioral analysis algorithms. To develop and test the algorithms a labeled dataset will be built.

Application Deadline: 12 October 2017

See Application Procedure Details


Closed Position

Econometric Behavioural Analysis through Machine Learning Techniques

The project aims at the study, design and development of Econometric and Machine Learning algorithms to be employed to understand and analyze the behaviors of users (eg. smart home assisted living, monitoring of security forces, visitors of cultural and natural sites) from heterogeneous data gathered with wearable devices (eg smart glasses, gesture control armband, gaze tracker). The models should be able to profile users with respect to a set of categories to be inferred automatically from the collected data and taking into account the cultural and natural domain. The developed models and algorithms have to anticipate and predict the behaviour of visitors (eg. interaction with objects) in order to recommend possible actions or choices to the user. The study also aims at the definition of proper evaluation measures to be used to understand the econometric impact of user/site manager choices.

Application Deadline: 12 October 2017

See Application Procedure Details