SPICE

semantic propositional image caption evaluation

Peter Anderson*, Basura Fernando, Mark Johnson, Stephen Gould

*Corresponding author for this work

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

158 Citations (Scopus)

Abstract

There is considerable interest in the task of automatically generating image captions. However, evaluation is challenging. Existing automatic evaluation metrics are primarily sensitive to n-gram overlap, which is neither necessary nor sufficient for the task of simulating human judgment. We hypothesize that semantic propositional content is an important component of human caption evaluation, and propose a new automated caption evaluation metric defined over scene graphs coined SPICE. Extensive evaluations across a range of models and datasets indicate that SPICE captures human judgments over model-generated captions better than other automatic metrics (e.g., system-level correlation of 0.88 with human judgments on the MS COCO dataset, versus 0.43 for CIDEr and 0.53 for METEOR). Furthermore, SPICE can answer questions such as which caption-generator best understands colors? and can caption-generators count?.

Original languageEnglish
Title of host publicationComputer Vision - 14th European Conference, ECCV 2016, Proceedings
EditorsBastian Leibe, Jiri Matas, Nicu Sebe, Max Welling
Place of PublicationCham, Switzerland
PublisherSpringer, Springer Nature
Pages382-398
Number of pages17
VolumePart V
ISBN (Print)9783319464534
DOIs
Publication statusPublished - 2016
EventEuropean Conference on Computer Vision (14th : 2016) - Amsterdam, Netherlands
Duration: 11 Oct 201614 Oct 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9909 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Conference

ConferenceEuropean Conference on Computer Vision (14th : 2016)
CountryNetherlands
CityAmsterdam
Period11/10/1614/10/16

Fingerprint Dive into the research topics of 'SPICE: semantic propositional image caption evaluation'. Together they form a unique fingerprint.

  • Cite this

    Anderson, P., Fernando, B., Johnson, M., & Gould, S. (2016). SPICE: semantic propositional image caption evaluation. In B. Leibe, J. Matas, N. Sebe, & M. Welling (Eds.), Computer Vision - 14th European Conference, ECCV 2016, Proceedings (Vol. Part V, pp. 382-398). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9909 LNCS). Cham, Switzerland: Springer, Springer Nature. https://doi.org/10.1007/978-3-319-46454-1_24