NanoSD：面向实时图像复原的边缘高效基础模型
NanoSD: Edge Efficient Foundation Model for Real Time Image Restoration
NanoSD 通过联合压缩 U-Net 和 VAE 编码器-解码器，在保持生成先验的同时实现边缘设备实时推理，为视觉任务提供可即插即用的轻量扩散基础模型，并揭示了参数减少与硬件效率不直接相关的设计原则。
原文依据 Existing lightweight variants predominantly compress the denoising U-Net or reduce the diffusion trajectory, which disrupts the underlying latent manifold and limits generalization beyond a single task.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论Stable Diffusion 1.5 等潜变量扩散模型提供了强大的生成先验，对图像复原极具价值，但其完整流程的计算量仍然过大，难以部署在边缘设备上。现有的轻量级变体主要压缩去噪 U-Net 或减少扩散轨迹，这会破坏潜在的潜变量流形，并限制其泛化到单一任务之外。我们提出 NanoSD，一系列通过联合对 U-Net 和变分自编码器编码器-解码器进行网络手术、特征级生成蒸馏和结构化架构缩放，从 Stable Diffusion 1.5 蒸馏得到的帕累托最优扩散基础模型。这种全流程协同设计保留了生成先验，同时产生的模型在准确率-延迟-尺寸前沿上占据不同的工作点（例如，130M-315M 参数，在移动级 NPU 上实现低至 20ms 的实时推理）。我们表明，仅减少参数并不与硬件效率相关，并提供了分析，揭示架构平衡、特征路由和潜空间保留如何共同塑造真实的设备端延迟。当作为即插即用骨干网络使用时，NanoSD 在图像超分辨率、图像去模糊、人脸复原和单目深度估计上实现了最优性能，在感知质量和实际可部署性上均优于先前的轻量级扩散模型。NanoSD 建立了一个通用扩散基础模型系列，适用于边缘设备上的实时视觉生成和复原。
英文原文摘要 1610 字符
Latent diffusion models such as Stable Diffusion 1.5 offer strong generative priors that are highly valuable for image restoration, yet their full pipelines remain too computationally heavy for deployment on edge devices. Existing lightweight variants predominantly compress the denoising U-Net or reduce the diffusion trajectory, which disrupts the underlying latent manifold and limits generalization beyond a single task. We introduce NanoSD, a family of Pareto-optimal diffusion foundation models distilled from Stable Diffusion 1.5 through network surgery, feature-wise generative distillation, and structured architectural scaling jointly applied to the U-Net and the VAE encoder-decoder. This full-pipeline co-design preserves the generative prior while producing models that occupy distinct operating points along the accuracy-latency-size frontier (e.g., 130M-315M parameters, achieving real-time inference down to 20ms on mobile-class NPUs). We show that parameter reduction alone does not correlate with hardware efficiency, and we provide an analysis revealing how architectural balance, feature routing, and latent-space preservation jointly shape true on-device latency. When used as a drop-in backbone, NanoSD enables state-of-the-art performance across image super-resolution, image deblurring, face restoration, and monocular depth estimation, outperforming prior lightweight diffusion models in both perceptual quality and practical deployability. NanoSD establishes a general-purpose diffusion foundation model family suitable for real-time visual generation and restoration on edge devices.