Visual question answering using external knowledge

Gulganjalli Narasimhan, Medhini

Permalink

https://hdl.handle.net/2142/104918 Copy

Description

Title

Visual question answering using external knowledge

Author(s)

Gulganjalli Narasimhan, Medhini

Issue Date

2019-04-25

Director of Research (if dissertation) or Advisor (if thesis)

• Schwing, Alexander G.
• Lazebnik, Svetlana

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Date of Ingest

2019-08-23T20:02:06Z

Keyword(s)

Visual question answering, knowledge bases, graph convolution networks

Abstract

Accurately answering a question about a given image requires combining observations with general knowledge. While this is effortless for humans, reasoning with general knowledge remains an algorithmic challenge. To advance research in this direction, a novel `fact-based' visual question answering (FVQA) task has been introduced recently along with a large set of curated facts which link two entities, i.e., two possible answers, via a relation. Given a question-image pair, keyword matching techniques have been employed to successively reduce the large set of facts and were shown to yield compelling results despite being vulnerable to misconceptions due to synonyms and homographs. To overcome these shortcomings, we introduce two new approaches in this work. We develop a learning-based approach which goes straight to the facts via a learned embedding space. We demonstrate state-of-the-art results on the challenging recently introduced factbased visual question answering dataset, outperforming competing methods by more than 5%. Upon further analysis, we observe that a successive process which considers one fact at a time to form a local decision is sub-optimal. To counter this, in our second approach we develop an entity graph and use a graph convolutional network to `reason' about the correct answer by jointly considering all entities. We show on the FVQA dataset that this leads to an improvement in accuracy of around 7% compared to the state-of-the-art.

Graduation Semester

2019-05

Type of Resource

text

Permalink

http://hdl.handle.net/2142/104918

Copyright and License Information

Copyright 2019 Medhini Gulganjalli Narasimhan

Owning Collections