3D scene graph generation from synthesis 3D indoor scene

Scene understanding in 3D vision has extended beyond object instance information to include high-level scene information, such as relationships between object instances. Scene graphs are a common representation of object relationships, but the long-tailed distribution of relationship types presents...

Full description

Bibliographic Details
Main Author: Qin, Huaiyuan
Other Authors: Jiang Xudong
Format: Final Year Project (FYP)
Language:English
Published: Nanyang Technological University 2023
Subjects:
Online Access:https://hdl.handle.net/10356/167002
_version_ 1824455715265183744
author Qin, Huaiyuan
author2 Jiang Xudong
author_facet Jiang Xudong
Qin, Huaiyuan
author_sort Qin, Huaiyuan
collection NTU
description Scene understanding in 3D vision has extended beyond object instance information to include high-level scene information, such as relationships between object instances. Scene graphs are a common representation of object relationships, but the long-tailed distribution of relationship types presents a challenge for accurate scene graph generation. Existing 3D indoor datasets focus mainly on object instance class and segmentation labels, making it difficult to be utilized to scene graph related tasks. In this FYP, we propose a synthesis 3D indoor dataset to collect data from virtual environments with both instance-level and predicate-level annotations. We also introduce a post-processing calibration method to handle the bias of long-tailed distribution in 3D scene graphs. Our experiment results show that the proposed method significantly improves the performance of the baseline model without changing its weights. We evaluate the proposed dataset and benchmark it on two 3D scene graph generation tasks, SGCls, and PredCls. This project contributes to the research in 3D vision and can benefit the fields of AR/VR and robotics.
first_indexed 2025-02-19T03:42:37Z
format Final Year Project (FYP)
id ntu-10356/167002
institution Nanyang Technological University
language English
last_indexed 2025-02-19T03:42:37Z
publishDate 2023
publisher Nanyang Technological University
record_format dspace
spelling ntu-10356/1670022023-07-07T17:37:28Z 3D scene graph generation from synthesis 3D indoor scene Qin, Huaiyuan Jiang Xudong School of Electrical and Electronic Engineering EXDJiang@ntu.edu.sg Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Scene understanding in 3D vision has extended beyond object instance information to include high-level scene information, such as relationships between object instances. Scene graphs are a common representation of object relationships, but the long-tailed distribution of relationship types presents a challenge for accurate scene graph generation. Existing 3D indoor datasets focus mainly on object instance class and segmentation labels, making it difficult to be utilized to scene graph related tasks. In this FYP, we propose a synthesis 3D indoor dataset to collect data from virtual environments with both instance-level and predicate-level annotations. We also introduce a post-processing calibration method to handle the bias of long-tailed distribution in 3D scene graphs. Our experiment results show that the proposed method significantly improves the performance of the baseline model without changing its weights. We evaluate the proposed dataset and benchmark it on two 3D scene graph generation tasks, SGCls, and PredCls. This project contributes to the research in 3D vision and can benefit the fields of AR/VR and robotics. Bachelor of Engineering (Electrical and Electronic Engineering) 2023-05-15T02:47:43Z 2023-05-15T02:47:43Z 2023 Final Year Project (FYP) Qin, H. (2023). 3D scene graph generation from synthesis 3D indoor scene. Final Year Project (FYP), Nanyang Technological University, Singapore. https://hdl.handle.net/10356/167002 https://hdl.handle.net/10356/167002 en A3110-221 application/pdf Nanyang Technological University
spellingShingle Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
Qin, Huaiyuan
3D scene graph generation from synthesis 3D indoor scene
title 3D scene graph generation from synthesis 3D indoor scene
title_full 3D scene graph generation from synthesis 3D indoor scene
title_fullStr 3D scene graph generation from synthesis 3D indoor scene
title_full_unstemmed 3D scene graph generation from synthesis 3D indoor scene
title_short 3D scene graph generation from synthesis 3D indoor scene
title_sort 3d scene graph generation from synthesis 3d indoor scene
topic Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
url https://hdl.handle.net/10356/167002
work_keys_str_mv AT qinhuaiyuan 3dscenegraphgenerationfromsynthesis3dindoorscene