Research map / Image generation / GANs
Multimodal Image Synthesis with Conditional Implicit Maximum Likelihood Estimation
Ke Li, Shichong Peng, Tianhao Zhang, Jitendra Malik
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 the noise encoder (ablation) | conditional IMLE layout synthesis · Cityscapes layout-to-image diversity | LPIPS: 0.10 -> 0.19 | Improved |
| add the loss rebalancing scheme (ablation) | conditional IMLE layout synthesis · Cityscapes layout-to-image diversity | LPIPS: 0.17 -> 0.19 | Improved |
| apply proposed rebalancing to CRN baseline (ablation) | CRN · Cityscapes layout-to-image diversity | LPIPS: 0.12 -> 0.13 | No clear effect |
| apply dataset rebalancing to Pix2pix-HD (ablation) | Pix2pix-HD · Cityscapes layout-to-image diversity | LPIPS | No clear effect |
| add noise input to CRN | CRN · Cityscapes layout-to-image diversity | LPIPS: 0.11 -> 0.12 | No clear effect |
| use conditional IMLE instead of CRN | CRN · Cityscapes layout-to-image diversity | LPIPS: 0.11 -> 0.19 | Improved |
| use conditional IMLE instead of CRN | CRN · Human evaluation | % images with more artifacts: 0.636 -> 0.364 | Improved |
| use SRIM instead of BicycleGAN | BicycleGAN · Multimodal super-resolution (sigma=0.3) | faithfulness-weighted variance: 4.31 x 10^-2 -> 5.91 x 10^-2 | Improved |
| use SRIM instead of MR-GAN | MR-GAN · Multimodal super-resolution (sigma=0.15) | faithfulness-weighted variance: 8.90 x 10^-5 -> 1.02 x 10^-3 | Improved |
| use RRDB generator in BicycleGAN | BicycleGAN · Multimodal super-resolution (sigma=0.3) | faithfulness-weighted variance: 4.31 x 10^-2 -> 3.00 x 10^-2 | Hurt |
Closest papers by meaning
- CHIMLE: Conditional Hierarchical IMLE for Multimodal Conditional Image Synthesis (NeurIPS 2022 · similarity 0.93)
- Harmonizing Maximum Likelihood with GANs for Multimodal Conditional Generation (ICLR 2019 · similarity 0.92)
- Diverse Image Synthesis From Semantic Layouts via Conditional IMLE (ICCV 2019 · similarity 0.90)
- Multimodal Image Synthesis and Editing: The Generative AI Era (TPAMI 2023 · similarity 0.88)
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 2024 · similarity 0.85)
- Training-free Dense-Aligned Diffusion Guidance for Modular Conditional Image Synthesis (CVPR 2025 · 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.
