Research map / Image generation / Autoregressive and latent generation
TIMI: Training-Free Image-to-3D Multi-Instance Generation with Spatial Fidelity
Xiao Cai, Pengpeng Zeng, Ji Zhang, Heng Tao Shen, Jingkuan Song, Lianli Gao
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 Instance-aware Separation Guidance (ISG) to Hunyuan3D 2.0 (ablation) | Hunyuan3D 2.0 · Synthetic 3D-Front (30 scenes) | CD-O: 0.0986 -> 0.0864 | Improved |
| add ISG without SGU (ablation) | Hunyuan3D 2.0 · Synthetic 3D-Front (30 scenes) | SSR: 0.697 -> 0.674 | Hurt |
| add SGU on top of ISG (ablation) | TIMI · Synthetic 3D-Front (30 scenes) | SSR: 0.674 -> 0.809 | Improved |
| add SGU on top of ISG (ablation) | TIMI · Synthetic 3D-Front (30 scenes) | LCD: 0.623 -> 0.598 | Improved |
| add geometry-adaptive modulation alone (ablation) | SGU · Synthetic 3D-Front (30 scenes) | LCD: 0.623 -> 0.597 | Improved |
| add spatial-stabilized regularization on top of GM (ablation) | SGU · Synthetic 3D-Front (30 scenes) | SSR: 0.764 -> 0.809 | Improved |
| guide cross-attention layers up to 8 instead of 4 (ablation) | ISG · Synthetic 3D-Front (30 scenes) | SSR: 0.809 -> 0.773 | Hurt |
| apply ISG for 25 steps instead of 15 (ablation) | ISG · Synthetic 3D-Front (30 scenes) | SSR: 0.809 -> 0.771 | Hurt |
| apply ISG for 10 steps instead of 15 (ablation) | ISG · Synthetic 3D-Front (30 scenes) | SSR: 0.809 -> 0.739 | Hurt |
| increase alpha from 0.1 to 0.5 (ablation) | SGU · Synthetic 3D-Front (30 scenes) | FS-O: 0.353 -> 0.314 | Hurt |
| increase sigma from 1.5 to 2.5 (ablation) | SGU · Synthetic 3D-Front (30 scenes) | SSR: 0.809 -> 0.782 | Hurt |
| reduce sigma from 1.5 to 0.5 (ablation) | SGU · Synthetic 3D-Front (30 scenes) | SSR: 0.809 -> 0.782 | Hurt |
Closest papers by meaning
- Ouroboros3D: Image-to-3D Generation via 3D-aware Recursive Diffusion (CVPR 2025 · similarity 0.86)
- PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (NeurIPS 2025 · similarity 0.86)
- InstanceDiffusion: Instance-level Control for Image Generation (CVPR 2024 · similarity 0.86)
- Disentangled 3D Scene Generation with Layout Learning (ICML 2024 · similarity 0.86)
- Direct3D-S2: Gigascale 3D Generation Made Easy with Spatial Sparse Attention (NeurIPS 2025 · similarity 0.86)
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.85)
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.
