Research map / 3D vision / Face and shape
High-Fidelity 3D Textured Shapes Generation by Sparse Encoding and Adversarial Decoding
Qi Zuo, Xiaodong Gu, Yuan Dong, Zhengyi Zhao, Weihao Yuan, Qiu Lingteng, Liefeng Bo, Zilong Dong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add sparse encoding of million-point clouds in place of 20K-point PointNet input (ablation) | Sparse3D reconstruction · reconstruction ablation | PSNR: 25.83 -> 26.11 | Improved |
| add sparse encoding (ablation) | Sparse3D reconstruction · reconstruction ablation | CD: 0.039 -> 0.036 | Improved |
| add adversarial finetuning to the baseline (ablation) | Sparse3D reconstruction · reconstruction ablation | PSNR: 25.83 -> 27.03 | Improved |
| add adversarial finetuning to the baseline (ablation) | Sparse3D reconstruction · reconstruction ablation | CD: 0.039 -> 0.039 | No clear effect |
| combine sparse encoding and adversarial finetuning (ablation) | Sparse3D reconstruction · reconstruction ablation | PSNR: 25.83 -> 27.59 | Improved |
| use Sparse3D instead of GET3D on Chair | single-class unconditional generation · ShapeNetV2 Chair | FID-3D: 33.18 -> 18.85 | Improved |
| use Sparse3D instead of GET3D on Car | single-class unconditional generation · ShapeNetV2 Car | FID-3D: 10.54 -> 10.17 | No clear effect |
| use Sparse3D instead of TexturedLAS for image-conditional generation | ShapeNetV2 multi-class · ShapeNetV2(S) | Chamfer distance: 0.149 -> 0.081 | Improved |
| train and evaluate on Objaverse instead of ShapeNetV2 | Sparse3D · image-conditional generation | PSNR: 20.03 -> 19.23 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GET3D — GET3D: A Generative Model of High Quality 3D Textured Shapes Learned from Images (NeurIPS 2022) · better on 6 of 6 shared results
- 3DILG — 3DILG: Irregular Latent Grids for 3D Generative Modeling (NeurIPS 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Convolutional Generation of Textured 3D Meshes (NeurIPS 2020 · similarity 0.90)
- Leveraging 2D Data to Learn Textured 3D Mesh Generation (CVPR 2020 · similarity 0.90)
- Learning Generative Models of Textured 3D Meshes From Real-World Images (ICCV 2021 · similarity 0.90)
- Fine Detailed Texture Learning for 3D Meshes With Generative Models (TPAMI 2023 · similarity 0.90)
- Make-A-Shape: a Ten-Million-scale 3D Shape Model (ICML 2024 · similarity 0.89)
- Structured 3D Latents for Scalable and Versatile 3D Generation (CVPR 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
