iCub Detecting Gazed Objects: A Pipeline Estimating Human Attention

TL;DR

This paper presents a visual-based system for detecting human gaze on objects in human-robot interactions, introduces a new dataset with over 22,000 images, and demonstrates the system's effectiveness on the iCub robot to enhance social and collaborative robotics.

Contribution

It introduces a novel gaze detection pipeline using visual feedback and provides a new benchmark dataset for human gaze estimation in tabletop HRI scenarios.

Findings

The system accurately detects gazed objects in real-time.

The new dataset enables benchmarking and improves gaze estimation methods.

Deployment on iCub demonstrates practical applicability in social robotics.

Abstract

This research report explores the role of eye gaze in human-robot interactions and proposes a learning system for detecting objects gazed at by humans using solely visual feedback. The system leverages face detection, human attention prediction, and online object detection, and it allows the robot to perceive and interpret human gaze accurately, paving the way for establishing joint attention with human partners. Additionally, a novel dataset collected with the humanoid robot iCub is introduced, comprising over 22,000 images from ten participants gazing at different annotated objects. This dataset serves as a benchmark for the field of human gaze estimation in table-top human-robot interaction (HRI) contexts. In this work, we use it to evaluate the performance of the proposed pipeline and examine the performance of each component. Furthermore, the developed system is deployed on the iCub, and a supplementary video showcases its functionality. The results demonstrate the potential of the proposed approach as a first step to enhance social awareness and responsiveness in social robotics, as well as improve assistance and support in collaborative scenarios, promoting efficient human-robot collaboration.