Research map / Point clouds / Shape and surface completion
DiT-3D: Exploring Plain Diffusion Transformers for 3D Shape Generation
Shentong Mo, Enze Xie, Ruihang Chu, Lanqing HONG, Matthias Nießner, Zhenguo Li
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 |
|---|---|---|---|
| Replace the 3D generation baseline with DiT-3D | ShapeNet Chair generation · ShapeNet Chair | 1-NNA CD: 53.70 -> 49.11 | Improved |
| Replace the 3D generation baseline with DiT-3D | ShapeNet Airplane generation · ShapeNet Airplane | 1-NNA CD: 66.44 -> 62.35 | Improved |
| Replace the 3D generation baseline with DiT-3D | ShapeNet Car generation · ShapeNet Car | COV EMD: 56.53 -> 56.38 | Hurt |
| Add voxelization and devoxelized prediction (ablation) | DiT-3D point-coordinate baseline · ShapeNet Chair | 1-NNA CD: 99.86 -> 67.46 | Improved |
| Add voxelization and devoxelized prediction (ablation) | DiT-3D point-coordinate baseline · ShapeNet Chair | COV CD: 7.768 -> 38.97 | Improved |
| Add 3D positional embeddings (ablation) | DiT-3D with voxel diffusion · ShapeNet Chair | 1-NNA CD: 67.46 -> 51.99 | Improved |
| Add 3D window attention (ablation) | DiT-3D with voxel diffusion and 3D positional embeddings · ShapeNet Chair | training cost (hours): 91.85 -> 41.67 | Improved |
| Add ImageNet pre-training (ablation) | DiT-3D · ShapeNet Chair | 1-NNA CD: 51.99 -> 49.07 | Improved |
| Use parameter-efficient fine-tuning of ImageNet-pretrained weights (ablation) | DiT-3D · ShapeNet Chair | 1-NNA CD: 49.07 -> 50.87 | Mixed |
| Transfer a source-class model using parameter-efficient fine-tuning (ablation) | DiT-3D, Airplane to Chair · ShapeNet Chair | 1-NNA CD: 51.99 -> 52.56 | Hurt |
| Transfer a source-class model using parameter-efficient fine-tuning (ablation) | DiT-3D, Chair to Airplane · ShapeNet Airplane | 1-NNA CD: 62.81 -> 63.58 | Hurt |
| Reduce patch size from 4 to 2 (ablation) | DiT-3D · ShapeNet Chair | 1-NNA CD: 51.99 -> 51.78 | Improved |
Later papers that compared with this method
5 later accepted papers report DiT-3D as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Rethinking 3D Shape Generation: Diffusion over Superquadrics (ICML 2026 · better on 1 of 6 shared results)
- Quartet of Diffusions: Structure-Aware Point Cloud Generation through Part and Symmetry Guidance (ICLR 2026 · better on 4 of 4 shared results)
- Scaling Diffusion Mamba with Bidirectional SSMs for Efficient 3D Shape Generation (AAAI 2025 · better on 11 of 11 shared results)
- LPCG: A Self-conditional Architecture for Labeled Point Cloud Generation (AAAI 2025 · better on 1 of 3 shared results)
- Fast Training of Diffusion Transformer with Extreme Masking for 3D Point Clouds Generation (ECCV 2024 · better on 8 of 8 shared results)
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.
- LION — LION: Latent Point Diffusion Models for 3D Shape Generation (NeurIPS 2022) · better on 6 of 8 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
- Fast Training of Diffusion Transformer with Extreme Masking for 3D Point Clouds Generation (ECCV 2024 · similarity 0.92)
- TIGER: Time-Varying Denoising Model for 3D Point Cloud Generation with Diffusion Process (CVPR 2024 · similarity 0.90)
- Scaling Diffusion Mamba with Bidirectional SSMs for Efficient 3D Shape Generation (AAAI 2025 · similarity 0.89)
- Neural Point Cloud Diffusion for Disentangled 3D Shape and Appearance Generation (CVPR 2024 · similarity 0.89)
- DiT4Edit: Diffusion Transformer for Image Editing (AAAI 2025 · similarity 0.88)
- PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (NeurIPS 2025 · similarity 0.87)
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.
