Does the agent say what it sees: an analysis in 3D question answering

The emerging research topic of 3D Question and Answering involves complex challenges such as semantic language understanding, 3D scene comprehension, and establishing correspondence between language and target objects. In this thesis, I conduct a comprehensive analysis on the state-of-the-art 3D Question Answering benchmark model, ScanQA, by comparing the effectiveness of various auxiliary tasks, identifying limitations in current evaluation methods, and conducting additional ablation studies on the visual component. The findings and potential future directions for the 3D Question Answering task are discussed to assist researchers in their ongoing investigations.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2023

Author dc:creator

• Zhang, Haomeng

Contributors dc:contributor

• Gui, Liangyan

• Multi-modal Learning
• 3d Question Answering

Statement dc:rights

• Copyright 2023 Haomeng Zhang

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/120420

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F120420