[论文] Gate3DGen: A Pre-processing Model that Selectively Purifies Real-World...

论文概要

研究领域: CV 作者: Xiaolin Li, Zihao Huang, Fanpeng Meng, Chen Yang, Guanjun Wu, Kai Zhang, Jiemin Fang, Qi Tian 发布时间: 2026-10-22 arXiv: 2610.31649

中文摘要

单图像 3D 生成在精选基准上已实现高保真度，但在光照不均、模糊和有遮挡的真实世界图像上性能急剧下降。净化此类输入需要决定何时干预、保留 3D 生成所需的视觉线索，同时让干净输入保持不变。我们提出 Gate3DGen——一个选择性预处理模型，作为门控器为现有 3D 生成器的图像输入把关，无需重新训练它们。退化感知路由器绕过被分类为干净的输入，并提取特征来指导其余输入的净化。以这些特征为条件，3D 定向单步净化器通过空间损失净化退化输入，鼓励保留物体身份、轮廓和纹理线索。为提供成对监督，我们构建了合成数据引擎来渲染 3D 资产的干净视图并合成复合退化。在合成基准上，净化器将 TRELLIS.2 网格距离相比退化输入降低了 61.5%，平均仅需 2.3 秒。VLM 评估和真实照片的人类偏好进一步支持了身份保留和退化校正能力。

原文摘要

Single-image 3D generation has achieved high fidelity on curated benchmarks, yet degrades sharply on real-world images with uneven illumination, blur, and occlusions. We introduce Gate3DGen, a selective pre-processing model that gates image inputs to existing 3D generators without retraining them. A degradation-aware router bypasses inputs classified as clean and extracts features to guide purification of the remaining inputs. Conditioned on these features, a 3D-oriented one-step purifier purifies degraded inputs with spatial losses that encourage preservation of object identity, contours, and textural cues. On the synthetic benchmark, the purifier reduces TRELLIS.2 mesh distance by 61.5% relative to degraded inputs. VLM evaluation and human preferences on real photographs further support ...

#论文 #arXiv #CV #小凯