Research map / Image generation / Autoregressive and latent generation
MixNMatch: Multifactor Disentanglement and Encoding for Conditional Image Generation
Check 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 MixNMatch instead of FineGAN | Unconditional image generation on CUB · CUB | FID: 11.25 -> 9.17 | Helped |
| Use MixNMatch instead of FineGAN | Unconditional image generation on Stanford Dogs · Stanford Dogs | FID: 25.66 -> 24.24 | Helped |
| Use MixNMatch instead of FineGAN | Unconditional image generation on Stanford Cars · Stanford Cars | FID: 16.03 -> 6.48 | Helped |
| Use MixNMatch instead of FineGAN | Unconditional image generation on CUB · CUB | Inception Score: 52.53 ± 0.45 -> 50.05 ± 0.75 | Hurt |
| Use MixNMatch instead of FineGAN | Unconditional image generation on Stanford Dogs · Stanford Dogs | Inception Score: 46.92 ± 0.61 -> 46.97 ± 0.51 | Helped |
| Use MixNMatch instead of FineGAN | Unconditional image generation on Stanford Cars · Stanford Cars | Inception Score: 32.62 ± 0.37 -> 31.12 ± 0.62 | Hurt |
| Use MixNMatch instead of FineGAN encoders | Fine-grained clustering · CUB, Stanford Dogs, Stanford Cars | NMI: 0.403 -> 0.422; 0.233 -> 0.324; 0.354 -> 0.357 | Helped |
| Use MixNMatch code mode instead of FineGAN | Shape and texture disentanglement on CUB · CUB | Shape keypoint L2-distance: 21.04 -> 20.57 | Helped |
| Use MixNMatch feature mode instead of code mode | Shape and texture disentanglement on CUB · CUB | Shape keypoint L2-distance: 20.57 -> 16.29 | Helped |
| Use MixNMatch feature mode instead of code mode | Shape and texture disentanglement on CUB · CUB | Texture color histogram χ²-distance: 0.540 -> 0.565 | Hurt |
| Add paired image-code adversarial loss (ablation) | MixNMatch code mode · CUB shape and texture disentanglement | Shape keypoint L2-distance: 60.41 -> 20.57 | Helped |
| Add code reprediction loss (ablation) | MixNMatch code mode · CUB shape and texture disentanglement | Shape keypoint L2-distance: 47.67 -> 20.57 | Helped |
Closest papers by meaning
- ContraGAN: Contrastive Learning for Conditional Image Generation (NeurIPS 2020 · similarity 0.89)
- Disentangled Image Generation Through Structured Noise Injection (CVPR 2020 · similarity 0.88)
- Diversity-Sensitive Conditional Generative Adversarial Networks (ICLR 2019 · similarity 0.87)
- Learning Disentangled Representation by Exploiting Pretrained Generative Models: A Contrastive Learning View (ICLR 2022 · similarity 0.87)
- Disentangled and Controllable Face Image Generation via 3D Imitative-Contrastive Learning (CVPR 2020 · similarity 0.87)
- Attack Deterministic Conditional Image Generative Models for Diverse and Controllable Generation (AAAI 2024 · similarity 0.86)
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.