Image Generation On Imagenet 32X32
评估指标
bpd
评测结果
各个模型在此基准测试上的表现结果
比较表格
| 模型名称 | bpd |
|---|---|
| nvae-a-deep-hierarchical-variational | 3.92 |
| glow-generative-flow-with-invertible-1x1 | 4.09 |
| mintnet-building-invertible-neural-networks | 4.06 |
| residual-flows-for-invertible-generative | 4.01 |
| variational-diffusion-models | 3.72 |
| generating-high-fidelity-images-with-subscale | 3.85 |
| stylegan-xl-scaling-stylegan-to-large-diverse | - |
| preventing-posterior-collapse-with-delta-vaes | 3.77 |
| pagoda-progressive-growing-of-a-one-step | - |
| very-deep-vaes-generalize-autoregressive-1 | 3.8 |
| pixel-recurrent-neural-networks | 3.86 |
| hierarchical-transformers-are-more-efficient | 3.74 |
| score-matching-model-for-unbounded-data-score-1 | 3.85 |
| improved-techniques-for-maximum-likelihood | 3.43 |
| multi-resolution-continuous-normalizing-flows | 3.77 |
| flow-improving-flow-based-generative-models | 3.86 |
| biva-a-very-deep-hierarchy-of-latent | 3.96 |
| reflected-diffusion-models | 3.74 |
| neural-diffusion-models | 3.55 |
| denoising-diffusion-probabilistic-models | 3.89 |
| self-supervised-gan-analysis-and-improvement-1 | - |
| on-maximum-likelihood-training-of-score-based | 3.76 |
| input-perturbation-reduces-exposure-bias-in | - |
| generative-modeling-with-bayesian-sample | 3.44 |
| image-transformer | 3.77 |
| flow-matching-for-generative-modeling | 3.53 |
| quasi-conservative-score-based-generative | - |
| augmented-normalizing-flows-bridging-the-gap | 3.92 |
| conditional-image-generation-with-pixelcnn | 3.83 |
| the-gan-is-dead-long-live-the-gan-a-modern | - |
| densely-connected-normalizing-flows | 3.63 |
| density-estimation-using-real-nvp | 4.28 |
| neural-flow-diffusion-models-learnable | 3.34 |