[论文] Sculpting NeRF Geometry: Human-Preference Fine-Tuning of a 3D-Aware Fa...

论文概要

研究领域: CV
作者: Archer Moore, Mingming Gong, Liam Hodgkinson
发布时间: 2026-06-25
arXiv: 2606.27305

中文摘要

基于人类反馈的强化学习（RLHF）用于 3D 生成已在多项工作中得到确立，但大多数现有流水线优化显式表面表征，通常通过将辐射场转换为网格并在表面监督数据上大量训练。相反，我们直接从辐射场密度（sigma）值的学习奖励来微调预训练的 3D 感知生成模型，无需外部提供的网格或形状先验。奖励模型无需预训练，在小规模偏好样本上易于训练，并在 3D 几何上产生鲁棒的改进。在无条件 3D 感知人脸 GAN（EG3D）上，我们的奖励直接读取神经辐射场（NeRF）的连续 3D 密度场，提供仅几何的学习信号，既不需要文本条件、网格提取，也不需要多视图渲染。密度一致性约束在几何重塑的同时保持 2D 外观的定性相似，以可测量但有界的分布代价（FID-50k 从 4.09 上升到 6.66）：作为概念验证，从单一标注者偏好训练的微调生成器在 74.4% 的成对比较中产生更受用户偏好的面部几何。

原文摘要

Reinforcement learning from human feedback (RLHF) for 3D generation is now established across a number of works, but most existing pipelines optimise explicit surface representations, often by converting radiance fields into meshes and training heavily on surface-supervised data. We instead fine-tune a pretrained 3D-aware generative model directly from a learned reward over radiance-field density (sigma) values, with no externally supplied mesh or shape prior. The reward model requires no pretraining, trains easily on a small set of preference samples, and yields robust improvement in 3D geometry. Working on an unconditional 3D-aware face GAN (EG3D), our reward reads the continuous 3D density field of the neural radiance field (NeRF) directly and supplies a geometry-only learning signal, r...

自动采集于 2026-06-28

#论文 #arXiv #CV #小凯