Research Project

SPECTACLE: From Spectators to Actors: Understanding Social Interactions through the Eyes of a Participant


UPC Project

Start Date


End Date


Project Code


Default project illustration


Project Description

This is a project within the Call for IRI MdM Internal Projects 2019, a call of R&D projects addressed to IRI early career researchers, under the María de Maeztu Unit of Excellence Programme.

The long-standing goal of building socially intelligent machines has driven research efforts to focus on the understanding of social interactions from videos, typically recorded by a fixed camera. This project aims at analyzing social interactions from videos captured by a camera worn by one of the interacting persons. One key advantage of the first-person point of view is that it allows to follow the wearer across different environments and capture fine details of interactions that would be missed by fixed cameras (e.g., non-verbal communication cues, such as gestures, etc.). However, it is not possible to observe the camera wearer directly, and his/her social behavior must be estimated, a challenging and so far often overlooked task.
In this context, we aim at determining the level of engagement of the camera wearer in a social event and classify the type of interaction. Such information will be very relevant for assistive and collaborative robotics developments in the upcoming years.

Project Publications

Conference Publications

  • S. Felicioni and M. Dimiccoli. Interaction-GCN: a Graph Convolutional Network based framework for social interaction recognition in egocentric videos, 28th IEEE International Conference on Image Processing, 2021, Anchorage, AK, USA, pp. 2348-2352.

    Open/Close abstract Abstract Info Info pdf PDF