Research map / 3D vision / 3D generation
Photo3D: Advancing Photorealistic 3D Generation through Structure-Aligned Detail Enhancement
Xinyue Liang, Zhiyuan Ma, Lingchen Sun, Yanjun Guo, Lei Zhang
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 |
|---|---|---|---|
| fine-tune Trellis on Photo3D-MV with realism loss | Trellis · ImageNet test | MUSIQ: 69.108 -> 72.385 | Improved |
| fine-tune Step1X-3D texturing on Photo3D-MV | Step1X-3D · ImageNet test | MUSIQ: 70.471 -> 72.215 | Improved |
| fine-tune TexGaussian on Photo3D-MV | TexGaussian · ImageNet test | MUSIQ: 65.003 -> 70.273 | Improved |
| fine-tune Trellis on Photo3D-MV | Trellis · ImageNet test | Gemini winning rate %: 68.1 -> 95.0 | Improved |
| fine-tune Trellis on Photo3D-MV | Trellis · Real 3D datasets | Human realism score: 3.9 -> 4.8 | Improved |
| fine-tune Step1X-3D on Photo3D-MV | Step1X-3D · Real 3D datasets | MANIQA: 0.416 -> 0.424 | Improved |
| fine-tune TexGaussian on Photo3D-MV | TexGaussian · Real 3D datasets | MANIQA: 0.441 -> 0.444 | No clear effect |
| add perceptual feature adaptation loss Ladapt (ablation) | Photo3D (Trellis) · ImageNet test | MANIQA: 0.308 -> 0.470 | Improved |
| add semantic structure matching loss Lmatch (ablation) | Photo3D (Trellis) · ImageNet test | MUSIQ: 71.069 -> 72.385 | Improved |
| use L2 loss instead of realism loss (ablation) | Photo3D (Trellis) · ImageNet test | MUSIQ: 72.385 -> 55.281 | Hurt |
| use Gram loss instead of realism loss (ablation) | Photo3D (Trellis) · ImageNet test | MUSIQ: 72.385 -> 69.646 | Hurt |
| use GAN loss instead of realism loss (ablation) | Photo3D (Trellis) · ImageNet test | MUSIQ: 72.385 -> 72.058 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TexGaussian — TexGaussian: Generating High-quality PBR Material via Octree-based 3D Gaussian Splatting (CVPR 2025) · better on 4 of 4 shared results
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
- Structured 3D Latents for Scalable and Versatile 3D Generation (CVPR 2025 · similarity 0.89)
- One-2-3-45++: Fast Single Image to 3D Objects with Consistent Multi-View Generation and 3D Diffusion (CVPR 2024 · similarity 0.88)
- PartGen: Part-level 3D Generation and Reconstruction with Multi-view Diffusion Models (CVPR 2025 · similarity 0.88)
- Interactive3D: Create What You Want by Interactive 3D Generation (CVPR 2024 · similarity 0.88)
- Cue3D: Quantifying the Role of Image Cues in Single-Image 3D Generation (NeurIPS 2025 · similarity 0.88)
- Hi3DEval: Advancing 3D Generation Evaluation with Hierarchical Validity (NeurIPS 2025 · similarity 0.88)
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.
