用于改善扩散模型收敛性的补偿采样
Compensation Sampling for Improved Convergence in Diffusion Models
AI总结:
针对扩散模型因初始重建误差累积导致质量下降、收敛缓慢的问题,本文提出以 U-Net 实现的补偿采样项,在多类人脸与自然图像任务上提升质量并将训练收敛加速最高十倍。
摘要:
扩散模型在图像生成中取得了卓越质量,但也付出了代价。迭代去噪需要许多时间步才能生成高保真图像。我们认为,去噪过程受到一个关键限制:由于对目标数据的初始重建不准确,重建误差会不断累积。这会导致输出质量降低以及收敛变慢。为解决该问题,我们提出补偿采样,以引导生成过程朝向目标域。我们引入一个以 U-Net 实现的补偿项,在训练期间以及可选的推理期间只增加可忽略的计算开销。我们的方法具有灵活性,并使用基准数据集 CIFAR-10、CelebA、CelebA-HQ、FFHQ-256 和 FSG,展示了其在无条件生成、人脸修复和人脸去遮挡中的应用。我们的方法在图像质量方面持续取得最先进结果,同时将训练期间去噪过程的收敛速度加快最高达一个数量级。
英文摘要:
Diffusion models achieve remarkable quality in image generation, but at a cost. Iterative denoising requires many time steps to produce high fidelity images. We argue that the denoising process is crucially limited by an accumulation of the reconstruction error due to an initial inaccurate reconstruction of the target data. This leads to lower quality outputs, and slower convergence. To address this issue, we propose compensation sampling to guide the generation towards the target domain. We introduce a compensation term, implemented as a U-Net, which adds negligible computation overhead during training and, optionally, inference. Our approach is flexible and we demonstrate its application in unconditional generation, face inpainting, and face de-occlusion using benchmark datasets CIFAR-10, CelebA, CelebA-HQ, FFHQ-256, and FSG. Our approach consistently yields state-of-the-art results in terms of image quality, while accelerating the denoising process to converge during training by up to an order of magnitude.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。