arXiv cs.AI· Elly Akhoundi, Mathieu Lamarre, Thibault Goyallon, Patrick Anderson, \'Etienne Danvoye·· 4 小时前AI 评分27
面向4K面部资产贴图的领域微调扩散超分辨率方法
Domain-Tuned Diffusion Super-Resolution for 4K Facial Asset Maps
AI 导读
研究者提出一种领域微调的扩散超分模型,可将面部资产贴图从1024分辨率4倍放大至4096。方法基于冻结的SDXL主干,按模态接入轻量LoRA模块,并用Tile ControlNet保留输入结构。在留出贴图上,albedo的LPIPS从0.21降至0.08,细节优于双线性及现成GAN与扩散超分方案,且避免位移贴图出现错位或过度浮雕。
正文 · 原文
Abstract:Physically based facial assets require 4K albedo and displacement maps for cinematic close-ups, yet generative pipelines commonly produce them at only 1K. Naively applying photographic super-resolution is a poor fit: facial UV maps have domain-specific statistics, and displacement is a scalar height field whose errors directly distort geometry. We present a domain-tuned diffusion upscaler for 4x (1024 to 4096) facial asset super-resolution. Starting from a bilinearly upsampled map, we adapt a frozen SDXL backbone per modality with lightweight LoRA modules, while a Tile ControlNet preserves input structure as the diffusion prior restores pore- and wrinkle-scale detail. On held-out maps, our method produces sharper, more faithful detail than bilinear and off-the-shelf GAN and diffusion super-resolvers, reducing albedo LPIPS from 0.21 to 0.08 while avoiding misplaced or excessive displacement relief. These results show that lightweight domain adaptation can produce render-ready 4K facial asset maps without a custom super-resolution architecture.
| Comments: | Accepted to SIGGRAPH Asia 2026 Technical Communications. 4 pages |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR) |
| Cite as: | arXiv:2611.21890 [cs.CV] |
| (or arXiv:2611.21890v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2611.21890 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Elly Akhoundi [view email]
[v1]
Mon, 16 Nov 2026 21:03:28 UTC (10,019 KB)
来源:arXiv cs.AI · arxiv.org