Extended Gaze Following: Detecting Objects in Videos Beyond the Camera Field of View

TL;DR

This paper introduces a novel approach for detecting and localizing objects in videos based solely on people's gaze directions, including objects outside the camera view, using a new spatial representation and deep learning models.

Contribution

It proposes a new top-view gaze representation, develops convolutional networks for object localization, and trains these models on synthetic data validated on real datasets.

Findings

Effective detection of objects outside camera view

Superior performance over heuristic methods

Robust localization using synthetic training data

Abstract

In this paper we address the problems of detecting objects of interest in a video and of estimating their locations, solely from the gaze directions of people present in the video. Objects can be indistinctly located inside or outside the camera field of view. We refer to this problem as extended gaze following. The contributions of the paper are the followings. First, we propose a novel spatial representation of the gaze directions adopting a top-view perspective. Second, we develop several convolutional encoder/decoder networks to predict object locations and compare them with heuristics and with classical learning-based approaches. Third, in order to train the proposed models, we generate a very large number of synthetic scenarios employing a probabilistic formulation. Finally, our methodology is empirically validated using a publicly available dataset.