Research map / Image generation / Text-to-image diffusion
PartCraft: Crafting Creative Objects by Parts
Kam Woh Ng, Xiatian Zhu, Yi-Zhe Song, Tao Xiang
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 the token bottleneck MLP encoder (ablation) | PartCraft · CUB-200-2011 | FID: 16.36 -> 12.86 | Improved |
| replace MSE attention loss with entropy-based normalized attention loss (ablation) | PartCraft · CUB-200-2011 part composition | EMR and CoSim | Improved |
| add bottleneck encoder to speed up learning (ablation) | PartCraft · convergence analysis | learning speed | Improved |
| increase lambda_attn from 0.01 to 0.1 (ablation) | PartCraft · CUB-200-2011 | EMR: 0.460 -> 0.339 | Hurt |
| decrease lambda_attn from 0.01 to 0.00001 (ablation) | PartCraft · CUB-200-2011 | FID: 12.86 -> 11.64 (EMR 0.460 -> 0.397) | Mixed |
| fine-tune the diffusion model (DreamBooth) instead of textual inversion only | Textual Inversion · CUB-200-2011 | EMR: 0.305 -> 0.355 | Improved |
| unconstrained fine-tuning of cross-attention K/V (Custom Diffusion) | Textual Inversion · CUB-200-2011 | FID: 10.10 -> 37.61 | Hurt |
| use PartCraft instead of Break-a-Scene | part reconstruction · CUB-200-2011 | EMR: 0.390 -> 0.460 | Improved |
Closest papers by meaning
- Composing Parts for Expressive Object Generation (CVPR 2025 · similarity 0.89)
- PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (NeurIPS 2025 · similarity 0.86)
- PartGen: Part-level 3D Generation and Reconstruction with Multi-view Diffusion Models (CVPR 2025 · similarity 0.85)
- AutoPartGen: Autoregressive 3D Part Generation and Discovery (NeurIPS 2025 · similarity 0.85)
- Dreamweaver: Learning Compositional World Models from Pixels (ICLR 2025 · similarity 0.84)
- SceneCraft: An LLM Agent for Synthesizing 3D Scenes as Blender Code (ICML 2024 · similarity 0.84)
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.
