Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HY3D-Bench: Generation of 3D Assets

Team Hunyuan3D, :, Bowen Zhang, Chunchao Guo, Dongyuan Guo, Haolin Liu, Hongyu Yan, Huiwen Shi, Jiaao Yu, Jiachen Xu, Jingwei Huang, Kunhong Li, Lifu Wang, Linus, Penghao Wang, Qingxiang Lin, Ruining Tang, Xianghui Yang, Yang Li, Yirui Guan, Yunfei Zhao, Yunhan Yang, Zeqiang Lai, Zhihao Liang, Zibo Zhao

TL;DR

HY3D-Bench addresses data bottlenecks in 3D generation by providing a high-quality, standardized data foundation. It delivers 252k training-ready assets, 240k part-level decompositions, and 125k synthetic long-tail assets via a full processing and AIGC synthesis pipeline. A standardized benchmark and evaluation protocol enables fair comparisons and reproducibility across 3D generation research. Validated by training a lightweight Hunyuan3D-2.1-Small model, HY3D-Bench demonstrates strong data quality benefits for 3D generation, perception, robotics, and digital content creation.

Abstract

While recent advances in neural representations and generative models have revolutionized 3D content creation, the field remains constrained by significant data processing bottlenecks. To address this, we introduce HY3D-Bench, an open-source ecosystem designed to establish a unified, high-quality foundation for 3D generation. Our contributions are threefold: (1) We curate a library of 250k high-fidelity 3D objects distilled from large-scale repositories, employing a rigorous pipeline to deliver training-ready artifacts, including watertight meshes and multi-view renderings; (2) We introduce structured part-level decomposition, providing the granularity essential for fine-grained perception and controllable editing; and (3) We bridge real-world distribution gaps via a scalable AIGC synthesis pipeline, contributing 125k synthetic assets to enhance diversity in long-tail categories. Validated empirically through the training of Hunyuan3D-2.1-Small, HY3D-Bench democratizes access to robust data resources, aiming to catalyze innovation across 3D perception, robotics, and digital content creation.

HY3D-Bench: Generation of 3D Assets

TL;DR

HY3D-Bench addresses data bottlenecks in 3D generation by providing a high-quality, standardized data foundation. It delivers 252k training-ready assets, 240k part-level decompositions, and 125k synthetic long-tail assets via a full processing and AIGC synthesis pipeline. A standardized benchmark and evaluation protocol enables fair comparisons and reproducibility across 3D generation research. Validated by training a lightweight Hunyuan3D-2.1-Small model, HY3D-Bench demonstrates strong data quality benefits for 3D generation, perception, robotics, and digital content creation.

Abstract

While recent advances in neural representations and generative models have revolutionized 3D content creation, the field remains constrained by significant data processing bottlenecks. To address this, we introduce HY3D-Bench, an open-source ecosystem designed to establish a unified, high-quality foundation for 3D generation. Our contributions are threefold: (1) We curate a library of 250k high-fidelity 3D objects distilled from large-scale repositories, employing a rigorous pipeline to deliver training-ready artifacts, including watertight meshes and multi-view renderings; (2) We introduce structured part-level decomposition, providing the granularity essential for fine-grained perception and controllable editing; and (3) We bridge real-world distribution gaps via a scalable AIGC synthesis pipeline, contributing 125k synthetic assets to enhance diversity in long-tail categories. Validated empirically through the training of Hunyuan3D-2.1-Small, HY3D-Bench democratizes access to robust data resources, aiming to catalyze innovation across 3D perception, robotics, and digital content creation.
Paper Structure (16 sections, 2 equations, 12 figures, 2 tables)

This paper contains 16 sections, 2 equations, 12 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. 3D Generation
  4. 3D Datasets
  5. Methods
  6. Hunyuan Objarverse
  7. Existing Enhanced Objaverse Dataset
  8. Full-level Data Processing
  9. Part-level Data Processing
  10. Synthetic Data Generation
  11. Data Distribution and Visualization
  12. Evaluation
  13. Implementation Details
  14. Experimental Results
  15. Conclusion
  16. ...and 1 more sections

Figures (12)

  • Figure 1: HY3D-Bench is a unified ecosystem for high-fidelity 3D content generation. Our framework introduces (a) 252k high-quality assets with watertight meshes and multi-view renderings, (b) 240k structured part-level decomposition enabling fine-grained control, and (c) AIGC-synthesized 125k long-tail category assets. This benchmark provides standardized training data and evaluation protocols for advancing 3D generation research.
  • Figure 2: The evolution of the 3D generation.
  • Figure 3: Full-level Data Processing Pipeline.
  • Figure 4: Part-level Data Processing Pipeline.
  • Figure 5: Synthetic Data Generating Pipeline.
  • ...and 7 more figures