Research map / 3D vision / 3D generation
Compositional Scene Representation Learning via Reconstruction: A Survey
Jinyang Yuan, Tonglin Chen, Bin Li, Xiangyang Xue
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 the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | AIR · Reconstruction average over 6 datasets (Test 1) | MSE: 3.27e-3 -> 1.87e-3 | Improved |
| use the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | GMIOO · Reconstruction average over 6 datasets (Test 1) | MSE: 3.76e-3 -> 1.49e-3 | Improved |
| use the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | Slot Attention · Reconstruction average over 6 datasets (Test 1) | MSE: 5.89e-3 -> 2.39e-3 | Improved |
| use the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | SPACE · Reconstruction average over 6 datasets (Test 2) | MSE: 11.34e-3 -> 5.08e-3 | Improved |
| use the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | N-EM · Reconstruction average over 6 datasets (Test 1) | MSE: 6.57e-3 -> 5.18e-3 | Improved |
| use the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | N-EM · Reconstruction average over 6 datasets (Test 2) | MSE: 12.41e-3 -> 13.84e-3 | Hurt |
| use the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | GENESIS-V2 · Reconstruction average over 6 datasets (Test 1) | MSE: 5.62e-3 -> 4.53e-3 | Improved |
| use the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | EfficientMORL · Reconstruction average over 6 datasets (Test 2) | MSE: 15.01e-3 -> 9.07e-3 | Improved |
| use the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | MONet · Reconstruction average over 6 datasets (Test 2) | MSE: 13.18e-3 -> 6.96e-3 | Improved |
| use the compositional multi-layer version instead of a single-layer or two-layer non-compositional version at equal representation length (ablation) | IODINE · Reconstruction average over 6 datasets (Test 2) | MSE: 8.60e-3 -> 4.55e-3 | Improved |
Closest papers by meaning
- Unsupervised Learning of Compositional Scene Representations from Multiple Unspecified Viewpoints (AAAI 2022 · similarity 0.91)
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 2025 · similarity 0.90)
- Towards a Unified Compositional Model for Visual Pattern Modeling (ICCV 2017 · similarity 0.89)
- Not Only Text: Exploring Compositionality of Visual Representations in Vision-Language Models (CVPR 2025 · similarity 0.87)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.87)
- Learning to Compose: Improving Object Centric Learning by Injecting Compositionality (ICLR 2024 · 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.
