Research map / Image generation / GANs
Diverse Conditional Image Generation by Stochastic Regression with Latent Drop-Out Codes
Yang He, Bernt Schiele, Mario Fritz
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 neighbors enhanced loss (ablation) | CRN · LFW diversity | identity embedding std: 1.098 -> 1.299 | Improved |
| add neighbors enhanced loss (ablation) | Ours (Separate) · LFW diversity | identity embedding std: 1.235 -> 1.427 | Improved |
| add neighbors enhanced loss (ablation) | Ours (Shared) · LFW diversity | identity embedding std: 1.198 -> 1.370 | Improved |
| add neighbors enhanced loss (ablation) | Ours (Shared) · LFW faces (mean of 10) | RMSE: 8.908 -> 8.956 | Hurt |
| add neighbors enhanced loss (ablation) | Ours (Separate) · LFW faces (best of 10) | RMSE: 5.528 -> 5.768 | Hurt |
| add neighbors enhanced loss (ablation) | CRN · LFW faces (mean of 10) | RMSE: 9.107 -> 8.612 | Improved |
| add neighbors enhanced loss (ablation) | Ours (Shared) · LFW landmarks (mean) | landmark error: 1.728 -> 1.864 | Hurt |
| add neighbors enhanced loss (ablation) | Ours (Shared) · LFW faces (mean) | SSIM: 0.701 -> 0.686 | Hurt |
| use early split instead of late split (CRN) (ablation) | stochastic regression · Oxford-IIIT Pet | max/mean distance diversity | Improved |
| use shared channel-wise dropout instead of separate branch parameters (ablation) | stochastic regression · Oxford-IIIT Pet | diversity | No clear effect |
| use channel-wise dropout instead of traditional dropout (ablation) | stochastic regression · Oxford-IIIT Pet | qualitative diversity | Improved |
| increase branches from 4 to 72 | Ours (Shared) · Oxford-IIIT Pet normals | mean normal error: 12.4 -> 11.7 | Improved |
Closest papers by meaning
- Diversity-Sensitive Conditional Generative Adversarial Networks (ICLR 2019 · similarity 0.89)
- Stochastic Conditional Generative Networks with Basis Decomposition (ICLR 2020 · similarity 0.88)
- Attack Deterministic Conditional Image Generative Models for Diverse and Controllable Generation (AAAI 2024 · similarity 0.88)
- Auto-Regressive Image Synthesis with Integrated Quantization (ECCV 2022 · similarity 0.87)
- Discovering Latent Graphs with GFlowNets for Diverse Conditional Image Generation (NeurIPS 2025 · similarity 0.86)
- Stochastic Interpolants via Conditional Dependent Coupling (NeurIPS 2026 · 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.
