Diffusion-GAN: Training GANs with Diffusion
Authors: Zhendong Wang, Huangjie Zheng, Pengcheng He, Weizhu Chen, Mingyuan Zhou
ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the advantages of Diffusion-GAN over strong GAN baselines on various datasets, showing that it can produce more realistic images with higher stability and data efficiency than state-of-the-art GANs. ... We conduct extensive experiments to answer the following questions: (a) Will Diffusion-GAN outperform state-of-the-art GAN baselines on benchmark datasets? (b) Will the diffusion-based noise injection help the learning of GANs in domain-agnostic tasks? (c) Will our method improve the performance of data-efficient GANs trained with a very limited amount of data? |
| Researcher Affiliation | Collaboration | 1The University of Texas at Austin, 2Microsoft Azure AI |
| Pseudocode | Yes | We provide the Diffusion-GAN algorithm in Algorithm 1. |
| Open Source Code | No | The paper does not provide an explicit statement or link indicating that the source code for the methodology is openly available. |
| Open Datasets | Yes | Datasets. We conduct experiments on image datasets ranging from low-resolution (e.g., 32 32) to high-resolution (e.g., 1024 1024) and from low-diversity to high-diversity: CIFAR-10 (Krizhevsky, 2009), STL-10 (Coates et al., 2011), LSUN-Bedroom (Yu et al., 2015), LSUN-Church (Yu et al., 2015), AFHQ(Cat/Dog/Wild) (Choi et al., 2020), and FFHQ (Karras et al., 2019). |
| Dataset Splits | No | The paper mentions using well-known datasets and their training subsets, but does not explicitly provide details about specific training/validation/test splits (e.g., percentages or exact counts for each split). |
| Hardware Specification | Yes | We run all our experiments with either 4 or 8 NVIDIA V100 GPUs depending on the demands of the inherited training configurations. |
| Software Dependencies | No | The paper mentions using base GAN implementations (Style GAN2, Projected GAN, Ins Gen) but does not provide specific version numbers for software dependencies like PyTorch, TensorFlow, or CUDA. |
| Experiment Setup | Yes | Diffusion config. For our diffusion-based noise injection, we set up a linearly increasing schedule for βt, where t {1, 2, . . . , T}. For pixel level injection in Style GAN2, we follow Ho et al. (2020b) and set β0 = 0.0001 and βT = 0.02. We adaptively modify T ranging from Tmin = 5 to Tmax = 1000. The image pixels are usually rescaled to [ 1, 1] so we set the Guassian noise standard deviation σ = 0.05. For feature level injection in Diffusion Projected GAN, we set β0 = 0.0001, βT = 0.01, Tmin = 5, Tmax = 500, and σ = 0.5. |