arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2502.03444cs.CVcs.AIcs.LG
收藏

Masked Autoencoders 是 Diffusion Models 的有效 Tokenizers

Masked Autoencoders Are Effective Tokenizers for Diffusion Models

Carnegie Mellon University(卡内基梅隆大学);AMD(超威半导体公司);Peking University(北京大学);William & Mary(威廉与玛丽学院)

Hao Chen, Yujin Han, Fangyi Chen, Xiang Li, Yidong Wang, Jindong Wang, Ze Wang, Zicheng Liu, Difan Zou, Bhiksha Raj

AI总结:

本文提出 MAETok,利用 mask 建模学习语义丰富的潜在空间,证明了潜在空间结构而非变分约束对 Diffusion Models 至关重要,在 ImageNet 上仅用 128 个 tokens 即实现 SOTA 性能并大幅提升训练与推理效率。

摘要:

近期 latent diffusion models 的进展证明了其在高分辨率图像合成方面的有效性。然而,为了更好地学习和生成扩散模型,来自 tokenizer 的潜在空间的属性仍未被充分探索。在理论和经验上,我们发现生成质量的提高与具有更好结构的潜在分布密切相关,例如具有更少 Gaussian Mixture 模式和更具判别性特征的分布。受这些见解的启发,我们提出了 MAETok,这是一种利用 mask 建模来学习语义丰富的潜在空间同时保持重建保真度的 autoencoder (AE)。大量实验验证了我们的分析,表明 autoencoders 的变分形式并非必要,仅使用来自 AE 的判别性潜在空间,在仅使用 128 个 tokens 的情况下,就能在 ImageNet 生成上实现最先进的性能。MAETok 取得了显著的实际改进,在 512x512 生成中实现了 1.69 的 gFID,训练速度提高了 76 倍,推理吞吐量提高了 31 倍。我们的研究结果表明,潜在空间的结构而非变分约束,对于有效的 diffusion models 至关重要。代码和训练好的模型已发布。

英文摘要:

Recent advances in latent diffusion models have demonstrated their effectiveness for high-resolution image synthesis. However, the properties of the latent space from tokenizer for better learning and generation of diffusion models remain under-explored. Theoretically and empirically, we find that improved generation quality is closely tied to the latent distributions with better structure, such as the ones with fewer Gaussian Mixture modes and more discriminative features. Motivated by these insights, we propose MAETok, an autoencoder (AE) leveraging mask modeling to learn semantically rich latent space while maintaining reconstruction fidelity. Extensive experiments validate our analysis, demonstrating that the variational form of autoencoders is not necessary, and a discriminative latent space from AE alone enables state-of-the-art performance on ImageNet generation using only 128 tokens. MAETok achieves significant practical improvements, enabling a gFID of 1.69 with 76x faster training and 31x higher inference throughput for 512x512 generation. Our findings show that the structure of the latent space, rather than variational constraints, is crucial for effective diffusion models. Code and trained models are released.

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。

↑