a list of current and past projects
This project presents a study on personal aesthetics, a recent soft biometrics application where the goal is to recognize
people by considering the images they like. We propose several soft biometric approaches on a dataset of 200 users,
40K images. Given a set of preferred image of a user, it extracts a set of features which are discriminative for his/her.
This project proposes new approaches, based on Computational Aesthetics, capable to infer the personality traits of Flickr users from
their galleries of favorite pictures. In particular, the approaches map
low-level features extracted from the pictures into numerical scores corresponding to the Big-Five Traits, both self-assessed and
attributed. The experiments were performed over 60,000 pictures tagged as favorite by 300 users (the PsychoFlickr Corpus).
The main idea was to create a tool able to stimulate the user to sense spatial depth from a tactile input. In particular we have created a table with four pressure sensors able to give four different sounds and haptic feedbacks
The general purpose of this project, in collaboration with Dr. Manganotti, was to analyze video sequences involving epileptic patients. In particular, we created a classifier able to discriminate different stages of an epileptic seizure. To face this problem we used some techniques based on Active Appearance Models (AAM).
Reconstruction of the 3D structure of a temple, knowing the number of points from 5 views of the corresponding 2D object.
An implementation of the Loy and Zelinski’s fast radial feature detector, revisited for 3D objects
Authorship attribution (AA) aims at recognizing automatically the author of a given text sample. Traditionally applied to literary texts, AA faces now the new challenge of recognizing the identity of people involved in chat conversations. These share many aspects with spoken conversations, but AA approaches did not take it into account so far. Hence, this thesis tries to fill the gap and proposes two novelties that improve the effectiveness of traditional AA approaches for this type of data: the first is to adopt features inspired by Conversation Analysis (in particular for turn-taking), the second is to extract the features from individual turns rather than from entire conversations. The experiments have been performed over a corpus of dyadic chat conversations (77 individuals in total). The performance in identifying the persons involved in each exchange, measured in terms of area under the Cumulative Match Characteristic curve, is 89.5%.
Extend the mElite game located in the space. You can negotiate objects, fuel, planets. For each asteroid you destroy, you earn money and life.
The project consisted in the implementation of a face recognition prototype using eigenfaces and the Qt library. The prototype is now working at the VIPS Lab as authentication system.
The aim was to use ARToolkit for designing 3D models. The program is able to recognize drawn and printed patterns (like a shape). We studied a mechanism to make the user interact with the program not only through a simple marker, but through keyboard or fingers motion
The purpose was to extract the blood vessels taking as input different retinal images