Research map / Image generation / Autoregressive and latent generation
Conditional Image Generation by Conditioning Variational Auto-Encoders
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the forward-KL IPA objective instead of the reverse-KL IPA-R objective (ablation) | IPA image completion · CIFAR-10 | LPIPS-GT: .0330 -> .0262 | Helped |
| Use the forward-KL IPA objective instead of the reverse-KL IPA-R objective (ablation) | IPA image completion · ImageNet-64 | LPIPS-GT: .166 -> .138 | Helped |
| Use the forward-KL IPA objective instead of the reverse-KL IPA-R objective (ablation) | IPA image completion · FFHQ-256 | FID: 8.82 -> 3.93 | Helped |
| Set sampling temperature to 0.85 (ablation) | IPA · CIFAR-10 | FID: 10.50 -> 8.61 | Helped |
| Set sampling temperature to 0.85 (ablation) | IPA · CIFAR-10 | LPIPS-GT: .0262 -> .0263 | No clear effect |
| Set sampling temperature to 0.85 (ablation) | IPA · ImageNet-64 | FID: 18.9 -> 15.1 | Helped |
| Set sampling temperature to 0.85 (ablation) | IPA · FFHQ-256 | P-IDS: 7.79 -> 8.50 | Helped |
| Use IPA rather than CoModGAN | Image completion · CIFAR-10 | LPIPS-GT: .0326 -> .0262 | Helped |
| Use IPA rather than CoModGAN | Image completion · ImageNet-64 | FID: 20.2 -> 18.9 | Helped |
| Use IPA rather than CoModGAN | Image completion · FFHQ-256 | FID: 2.33 -> 3.93 | Hurt |
| Use pretrained weights rather than train the conditional VAE from scratch | IPA · CIFAR-10 | FID | Helped |
| Use ImageNet-pretrained weights rather than train from scratch | IPA · ImageNet-64 | FID | Helped |
Closest papers by meaning
- IntroVAE: Introspective Variational Autoencoders for Photographic Image Synthesis (NeurIPS 2018 · similarity 0.88)
- Harmonizing Maximum Likelihood with GANs for Multimodal Conditional Generation (ICLR 2019 · similarity 0.87)
- Generating Diverse High-Fidelity Images with VQ-VAE-2 (NeurIPS 2019 · similarity 0.87)
- Adversarial Symmetric Variational Autoencoder (NeurIPS 2017 · similarity 0.86)
- Autoencoding Conditional Neural Processes for Representation Learning (ICML 2024 · similarity 0.86)
- A Simple Approach to Unifying Diffusion-based Conditional Generation (ICLR 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.