Research map / Image generation / GANs
LR-GAN: Layered Recursive Generative Adversarial Networks for Image Generation
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 layered recursive generation with foreground appearance, shape, and pose instead of DCGAN generation | CIFAR-10 image generation · CIFAR-10 | Inception Score (pre-trained Inception net): 6.64±0.14 -> 7.17±0.07 | Helped |
| Use layered recursive generation with foreground appearance, shape, and pose instead of DCGAN generation | CIFAR-10 image generation · CIFAR-10 | Adversarial Accuracy: 37.81±0.02 -> 44.22 ± 0.08 | Helped |
| Use layered recursive generation with foreground appearance, shape, and pose instead of DCGAN generation | CIFAR-10 image generation · CIFAR-10 | Adversarial Divergence: 7.58±0.04 -> 5.57±0.06 | Helped |
| Use layered recursive generation with foreground appearance, shape, and pose instead of DCGAN generation | MNIST-ONE image generation · MNIST-ONE | Inception Score (target-dataset classifier): 6.42±0.03 -> 7.15±0.04 | Helped |
| Use layered recursive generation with foreground appearance, shape, and pose instead of DCGAN generation | MNIST-ONE image generation · MNIST-ONE | Adversarial Accuracy: 26.12 ± 0.07 -> 26.61 ± 0.06 | Helped |
| Use layered recursive generation with foreground appearance, shape, and pose instead of DCGAN generation | CUB-200 image generation · CUB-200 | human preference: 31.6% -> 68.4% | Helped |
| Use layered recursive generation with foreground appearance, shape, and pose instead of DCGAN generation | CIFAR-10 image generation · CIFAR-10 | non-recognizability rate: 77.7% -> 67.3% | Helped |
| Add affine transformations to foreground layers (ablation) | LR-GAN · MNIST-ONE, CUB-200, CIFAR-10 | qualitative generation quality | Helped |
| Add a mask generator for foreground shapes (ablation) | LR-GAN · CUB-200, CIFAR-10 | plausible image generation | Helped |
| Add a mask generator for foreground shapes (ablation) | LR-GAN · CUB-200 | training convergence | Helped |
Closest papers by meaning
- Constrained Generative Adversarial Networks for Interactive Image Generation (CVPR 2019 · similarity 0.89)
- A Layer-Based Sequential Framework for Scene Generation with GANs (AAAI 2019 · similarity 0.89)
- DeLiGAN : Generative Adversarial Networks for Diverse and Limited Data (CVPR 2017 · similarity 0.89)
- Cluster-Guided Image Synthesis With Unconditional Models (CVPR 2022 · similarity 0.89)
- Generative Adversarial Transformers (ICML 2021 · similarity 0.88)
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · similarity 0.88)
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.