可扩展的高分辨率像素空间图像合成与Hourglass扩散变换器
Scalable High-Resolution Pixel-Space Image Synthesis with Hourglass Diffusion Transformers
AI总结：
本文提出Hourglass Diffusion Transformer，通过线性扩展像素数量实现高分辨率图像生成，无需传统高分辨率训练技术， 在ImageNet和FFHQ数据集上取得新突破。
摘要：
我们提出Hourglass Diffusion Transformer (HDiT)，一种具有线性像素计数扩展能力的图像生成模型，支持在像素空间中直接训练高分辨率图像（例如1024×1024）。基于Transformer架构，该模型结合了卷积U-Net的效率和Transformer的可扩展性。HDiT无需多尺度架构、潜在自编码器或自条件等传统高分辨率训练技术即可成功训练。我们证明HDiT在ImageNet 256²上与现有模型竞争， 并在FFHQ-1024²上为扩散模型设定了新的状态-of-the-art。
英文摘要：
We present the Hourglass Diffusion Transformer (HDiT), an image generative model that exhibits linear scaling with pixel count, supporting training at high-resolution (e.g. $1024 \times 1024$) directly in pixel-space. Building on the Transformer architecture, which is known to scale to billions of parameters, it bridges the gap between the efficiency of convolutional U-Nets and the scalability of Transformers. HDiT trains successfully without typical high-resolution training techniques such as multiscale architectures, latent autoencoders or self-conditioning. We demonstrate that HDiT performs competitively with existing models on ImageNet $256^2$, and sets a new state-of-the-art for diffusion models on FFHQ-$1024^2$.
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。
