高效且有效的掩碼圖像生成模型

Effective and Efficient Masked Image Generation Models

March 10, 2025

作者: Zebin You, Jingyang Ou, Xiaolu Zhang, Jun Hu, Jun Zhou, Chongxuan Li

cs.AI

摘要

尽管掩码图像生成模型和掩码扩散模型在设计动机和目标上有所不同，但我们观察到它们可以在一个统一的框架下进行整合。基于这一洞见，我们深入探索了训练和采样的设计空间，识别出对性能和效率均有贡献的关键因素。在此探索过程中，我们根据观察到的改进，开发了我们的模型，称为eMIGM。实验表明，eMIGM在ImageNet生成任务上表现出色，通过Fr\'echet Inception Distance (FID) 衡量。特别是在ImageNet 256x256数据集上，在相似的功能评估次数（NFEs）和模型参数数量下，eMIGM超越了开创性的VAR模型。此外，随着NFE和模型参数的增加，eMIGM在仅需不到40%的NFE的情况下，实现了与最先进的连续扩散模型相当的性能。同时，在ImageNet 512x512数据集上，仅需约60%的NFE，eMIGM便超越了当前最先进的连续扩散模型。

English

Although masked image generation models and masked diffusion models are designed with different motivations and objectives, we observe that they can be unified within a single framework. Building upon this insight, we carefully explore the design space of training and sampling, identifying key factors that contribute to both performance and efficiency. Based on the improvements observed during this exploration, we develop our model, referred to as eMIGM. Empirically, eMIGM demonstrates strong performance on ImageNet generation, as measured by Fr\'echet Inception Distance (FID). In particular, on ImageNet 256x256, with similar number of function evaluations (NFEs) and model parameters, eMIGM outperforms the seminal VAR. Moreover, as NFE and model parameters increase, eMIGM achieves performance comparable to the state-of-the-art continuous diffusion models while requiring less than 40% of the NFE. Additionally, on ImageNet 512x512, with only about 60% of the NFE, eMIGM outperforms the state-of-the-art continuous diffusion models.