Research map / 3D vision / Face and shape
Bottom-up modeling of repeated elements via single image analysis-by-synthesis
syrine kalleli, Alexei Efros, Mathieu Aubry
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 appearance head (ablation) | Ours · FSC-147 subset | image reconstruction SSIM: 0.606 -> 0.737 | Improved |
| add appearance head (ablation) | Ours · FSC-147 subset | image reconstruction LPIPS: 0.384 -> 0.242 | Improved |
| add appearance head (ablation) | Ours · FSC-147 subset | counting MAE: 14.99 -> 14.11 | Improved |
| add sparsity loss (ablation) | Ours · FSC-147 subset | counting MAE: 15.62 -> 14.11 | Improved |
| add sparsity loss (ablation) | Ours · FSC-147 subset | image reconstruction SSIM: 0.689 -> 0.737 | Improved |
| add pretraining and crop-based training (vs direct optimization) (ablation) | Ours · FSC-147 subset | counting MAE: 27.54 -> 14.11 | Improved |
| add synthetic pretraining (ablation) | Ours · FSC-147 subset | counting MAE: 18.81 -> 14.11 | Improved |
| add crop-based optimization (ablation) | Ours · FSC-147 subset | image reconstruction PSNR: 17.45 -> 23.76 | Improved |
| add full image refinement (ablation) | Ours · FSC-147 subset | image reconstruction SSIM: 0.576 -> 0.737 | Improved |
| add full image refinement (ablation) | Ours · FSC-147 subset | counting MAE: 14.11 -> 14.11 | No clear effect |
| use supervised box/count signal in pretraining (ablation) | Ours · FSC-147 subset | counting MAE: 15.48 -> 14.11 | No clear effect |
| add exemplar loss L_ex (ablation) | Ours · FSC-147 subset | prototype SSIM: 0.700 -> 0.751 | Improved |
Closest papers by meaning
- Unsupervised Layered Image Decomposition Into Object Prototypes (ICCV 2021 · similarity 0.86)
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · similarity 0.84)
- Unlocking Compositional Generalization in Continual Few-Shot Learning (NeurIPS 2026 · similarity 0.84)
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 2025 · similarity 0.84)
- Reconstruct and Match: Out-of-Distribution Robustness via Topological Homogeneity (NeurIPS 2024 · similarity 0.84)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · 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.
