Research map / Image generation / Autoregressive and latent generation
Inducing Hierarchical Compositional Model by Sparsifying Generator Network
Xianglei Xing, Tianfu Wu, Song-Chun Zhu, Ying Nian Wu
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 sparse activation to induce the AND-OR generator | AND-OR model · CelebA | FID: 18.85 -> 16.62 | Improved |
| use sparse activation to induce the AND-OR generator | AND-OR model · HumanFashion | FID: 10.19 -> 8.65 | Improved |
| use sparse activation to induce the AND-OR generator | AND-OR model · Stanford cars | FID: 31.62 -> 28.36 | Improved |
| use sparse activation to induce the AND-OR generator | AND-OR model · LSUN bedrooms | FID: 33.81 -> 29.70 | Improved |
| use the proposed generator for reconstruction | AND-OR model · CelebA | per-pixel MSE: 0.016 -> 0.011 | Improved |
| use the proposed generator for reconstruction | AND-OR model · HumanFashion | per-pixel MSE: 0.033 -> 0.024 | Improved |
| use the proposed generator for reconstruction | AND-OR model · Stanford cars | per-pixel MSE: 0.078 -> 0.054 | Improved |
| use the proposed generator for reconstruction | AND-OR model · LSUN bedrooms | per-pixel MSE: 0.127 -> 0.097 | Improved |
| add sparsity-inducing constraints (ablation) | generator network · CelebA | template matching score: 0.33 -> 0.83 | Improved |
| add sparsity-inducing constraints (ablation) | generator network · HumanFashion | template matching score: 0.29 -> 0.81 | Improved |
| add sparsity-inducing constraints (ablation) | generator network · Cars | template matching score: 0.31 -> 0.76 | Improved |
| add sparsity-inducing constraints (ablation) | generator network · Bedroom | template matching score: 0.23 -> 0.72 | Improved |
Closest papers by meaning
- Greedy Structure Learning of Hierarchical Compositional Models (CVPR 2019 · similarity 0.88)
- Generative Hierarchical Learning of Sparse FRAME Models (CVPR 2017 · similarity 0.87)
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 2025 · similarity 0.86)
- Cooperative Training of Descriptor and Generator Networks (TPAMI 2018 · similarity 0.86)
- Compositional Scene Representation Learning via Reconstruction: A Survey (TPAMI 2023 · similarity 0.86)
- How Deep Networks Learn Sparse and Hierarchical Data: the Sparse Random Hierarchy Model (ICML 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.
