File Download
Supplementary

Conference Paper: GPT4Point: A Unified Framework for Point-Language Understanding and Generation

TitleGPT4Point: A Unified Framework for Point-Language Understanding and Generation
Authors
Issue Date1-May-2024
Abstract

Multimodal Large Language Models (MLLMs) have excelled in 2D image-text comprehension and image generation, but their understanding of the 3D world is notably deficient, limiting progress in 3D language understanding and generation. To solve this problem, we introduce GPT4Point, an innovative groundbreaking point-language multimodal model designed specifically for unified 3D object understanding and generation within the MLLM framework. GPT4Point as a powerful 3D MLLM seamlessly can execute a variety of point-text reference tasks such as point-cloud captioning and Q&A. Additionally, GPT4Point is equipped with advanced capabilities for controllable 3D generation, it can get high-quality results through a low-quality point-text feature maintaining the geometric shapes and colors. To support the expansive needs of 3D object-text pairs, we develop Pyramid-XL, a point-language dataset annotation engine. It constructs a large-scale database over 1M objects of varied text granularity levels from the Objaverse-XL dataset, essential for training GPT4Point. A comprehensive benchmark has been proposed to evaluate 3D point-language understanding capabilities. In extensive evaluations, GPT4Point has demonstrated superior performance in understanding and generation.


Persistent Identifierhttp://hdl.handle.net/10722/346131

 

DC FieldValueLanguage
dc.contributor.authorQi, Zhangyang-
dc.contributor.authorFang, Ye-
dc.contributor.authorSun, Zeyi-
dc.contributor.authorWu, Xiaoyang-
dc.contributor.authorWu, Tong-
dc.contributor.authorWang, Jiaqi-
dc.contributor.authorLin, Dahua-
dc.contributor.authorZhao, Hengshuang-
dc.date.accessioned2024-09-10T00:30:42Z-
dc.date.available2024-09-10T00:30:42Z-
dc.date.issued2024-05-01-
dc.identifier.urihttp://hdl.handle.net/10722/346131-
dc.description.abstract<p>Multimodal Large Language Models (MLLMs) have excelled in 2D image-text comprehension and image generation, but their understanding of the 3D world is notably deficient, limiting progress in 3D language understanding and generation. To solve this problem, we introduce GPT4Point, an innovative groundbreaking point-language multimodal model designed specifically for <strong>unified 3D object understanding and generation within the MLLM framework</strong>. GPT4Point as a powerful 3D MLLM seamlessly can execute a variety of point-text reference tasks such as point-cloud captioning and Q&A. Additionally, GPT4Point is equipped with advanced capabilities for controllable 3D generation, it can get high-quality results through a low-quality point-text feature maintaining the geometric shapes and colors. To support the expansive needs of 3D object-text pairs, we develop <strong>Pyramid-XL, a point-language dataset annotation engine</strong>. It constructs a large-scale database over 1M objects of varied text granularity levels from the Objaverse-XL dataset, essential for training GPT4Point. <strong>A comprehensive benchmark</strong> has been proposed to evaluate 3D point-language understanding capabilities. In extensive evaluations, GPT4Point has demonstrated superior performance in understanding and generation.<br></p>-
dc.languageeng-
dc.relation.ispartof2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (17/06/2024-21/06/2024, Seattle)-
dc.titleGPT4Point: A Unified Framework for Point-Language Understanding and Generation-
dc.typeConference_Paper-
dc.description.naturepublished_or_final_version-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats