Interpretable feature maps for robot attention

Kasim Terzić*, J. M.H. du Buf

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Downloads (Pure)

Abstract

Attention is crucial for autonomous agents interacting with complex environments. In a real scenario, our expectations drive attention, as we look for crucial objects to complete our understanding of the scene. But most visual attention models to date are designed to drive attention in a bottom-up fashion, without context, and the features they use are not always suitable for driving top-down attention. In this paper, we present an attentional mechanism based on semantically meaningful, interpretable features. We show how to generate a low-level semantic representation of the scene in real time, which can be used to search for objects based on specific features such as colour, shape, orientation, speed, and texture.

Original languageEnglish
Title of host publicationUniversal Access in Human–Computer Interaction. Design and Development Approaches and Methods
Subtitle of host publication11th International Conference, UAHCI 2017, Held as Part of HCI International 2017, Vancouver, BC, Canada, July 9–14, 2017, Proceedings, Part I
EditorsMargherita Antona, Constantine Stephanidis
Place of PublicationCham
PublisherSpringer
Pages456-467
Number of pages12
ISBN (Electronic)9783319587066
ISBN (Print)9783319587059
DOIs
Publication statusPublished - 2017
Event11th International Conference on Universal Access in Human-Computer Interaction, UAHCI 2017, held as part of the 19th International Conference on Human-Computer Interaction, HCI 2017 - Vancouver, Canada
Duration: 9 Jul 201714 Jul 2017
Conference number: 11
http://2017.hci.international/index.php

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10277
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference11th International Conference on Universal Access in Human-Computer Interaction, UAHCI 2017, held as part of the 19th International Conference on Human-Computer Interaction, HCI 2017
Abbreviated titleUAHCI
Country/TerritoryCanada
CityVancouver
Period9/07/1714/07/17
Internet address

Fingerprint

Dive into the research topics of 'Interpretable feature maps for robot attention'. Together they form a unique fingerprint.

Cite this