Research map / Image generation / GANs
Multimodal Conditional Image Synthesis with Product-of-Experts GANs
Xun Huang, Arun Mallya, Ting-Chun Wang, Ming-Yu Liu
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 product-of-experts generator instead of concatenation with modality dropout (ablation) | PoE-GAN · MM-CelebA-HQ 256x256 (unconditional) | FID: 29.3 -> 14.9 | Improved |
| use product-of-experts generator instead of concatenation with modality dropout (ablation) | PoE-GAN · MS-COCO 2017 64x64 (unconditional) | FID: 59.1 -> 26.6 | Improved |
| add KL loss (ablation) | PoE-GAN · MM-CelebA-HQ 256x256 (unconditional) | FID: 29.1 -> 14.9 | Improved |
| add KL loss (ablation) | PoE-GAN · MS-COCO 2017 64x64 (unconditional) | FID: 59.3 -> 26.6 | Improved |
| add modality dropout (ablation) | PoE-GAN · MM-CelebA-HQ 256x256 (unconditional) | FID: 30.8 -> 14.9 | Improved |
| add modality dropout (ablation) | PoE-GAN · MS-COCO 2017 64x64 (unconditional) | FID: 86.2 -> 26.6 | Improved |
| add multiscale multimodal projection discriminator (MMPD) instead of concatenation discriminator (ablation) | PoE-GAN · MM-CelebA-HQ 256x256 (unconditional) | FID: 21.5 -> 14.9 | Improved |
| add multiscale multimodal projection discriminator (MMPD) instead of concatenation discriminator (ablation) | PoE-GAN · MS-COCO 2017 64x64 (unconditional) | FID: 43.1 -> 26.6 | Improved |
| add image contrastive loss (ablation) | PoE-GAN · MM-CelebA-HQ 256x256 (unconditional) | FID: 15.4 -> 14.9 | Mixed |
| add image contrastive loss (ablation) | PoE-GAN · MS-COCO 2017 64x64 (unconditional) | FID: 25.7 -> 26.6 | Hurt |
| add text contrastive loss (ablation) | PoE-GAN · MM-CelebA-HQ 256x256 (text) | FID: 15.0 -> 13.7 | Improved |
| add text contrastive loss (ablation) | PoE-GAN · MS-COCO 2017 64x64 (text) | FID: 26.0 -> 22.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- TediGAN — TediGAN: Text-Guided Diverse Face Image Generation and Manipulation (CVPR 2021) · better on 4 of 4 shared results
- DF-GAN — DF-GAN: A Simple and Effective Baseline for Text-to-Image Synthesis (CVPR 2022) · better on 1 of 1 shared result
- OASIS — OASIS: Only Adversarial Supervision for Semantic Image Synthesis (IJCV 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Multimodal Image Synthesis with Conditional Implicit Maximum Likelihood Estimation (IJCV 2020 · similarity 0.88)
- Conditional Image Synthesis with Auxiliary Classifier GANs (ICML 2017 · similarity 0.87)
- Harmonizing Maximum Likelihood with GANs for Multimodal Conditional Generation (ICLR 2019 · similarity 0.87)
- High-Resolution Image Synthesis and Semantic Manipulation With Conditional GANs (CVPR 2018 · similarity 0.87)
- Exploring Sparse MoE in GANs for Text-conditioned Image Synthesis (CVPR 2025 · similarity 0.87)
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 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.
