Research map / 3D vision / 3D generation
ROOTS: Object-Centric Representation and Rendering of 3D Scenes
Chang Chen, Fei Deng, Sungjin Ahn
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 |
|---|---|---|---|
| use object-centric 3D models (ROOTS) instead of scene-level GQN | GQN · 1-3 Shapes | MSE: 40.62 -> 30.60 | Improved |
| use object-centric 3D models instead of scene-level GQN | GQN · 3-5 Shapes | MSE: 62.73 -> 54.45 | Improved |
| use object-centric 3D models instead of scene-level GQN | GQN · Multi-Shepard-Metzler | MSE: 46.22 -> 42.22 | Improved |
| use object-centric 3D models instead of scene-level GQN | GQN · ShapeNet arrangement | MSE: 301.62 -> 212.77 | Improved |
| use ROOTS 3D segmentation instead of 2D IODINE | IODINE · 1-3 Shapes | ARI: 0.8217 -> 0.9477 | Improved |
| use ROOTS instead of IODINE on Multi-Shepard-Metzler | IODINE · Multi-Shepard-Metzler | ARI: Failed -> 0.9303 | Improved |
| use ROOTS instead of GQN when test scenes have fewer objects than training | GQN · 1-3 Shapes (trained on 2-4 Shapes) | MSE: 30.35 -> 24.27 | Improved |
| use ROOTS instead of GQN when test scenes have more objects than training | GQN · 3-5 Shapes (trained on 2-4 Shapes) | MSE: 86.68 -> 67.98 | Improved |
| use ROOTS representation instead of GQN for retrieve object task | GQN · Retrieve Object 3D | accuracy: 81.31% -> 90.38% | Improved |
| use ROOTS representation instead of GQN for pair-wise relational reasoning | GQN · Find Pair | accuracy: 12.48% -> 84.70% | Improved |
| add more context observations | ROOTS · Shapes counting and AP | object counting accuracy and AP | Improved |
| add object-attention grouping before inferring object appearance in the encoder (ablation) | ROOTS encoder · Shapes | trainability: failed to decompose scenes -> trainable | Improved |
Closest papers by meaning
- Object Scene Representation Transformer (NeurIPS 2022 · similarity 0.90)
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 2025 · similarity 0.88)
- PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (NeurIPS 2025 · similarity 0.87)
- Orientation Matters: Making 3D Generative Models Orientation-Aligned (NeurIPS 2025 · similarity 0.87)
- LT3SD: Latent Trees for 3D Scene Diffusion (CVPR 2025 · similarity 0.87)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 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.
