用后训练提升图像修复效果，解决过平滑和泛化差问题
IRPO: Boosting Image Restoration via Post-training GRPO
- 选取30%表现差样本进行后训练，提升效率与精度
- 融合结构保真、视觉质量与任务特定奖励，优化修复效果
- 在域内和域外任务上分别提升0.93dB和3.43dB，适合图像修复研究者
后训练在高层生成中已证明有效，但在低层视觉中仍研究不足。现有图像修复方法常依赖固定像素级拟合真实图像，易导致过平滑且泛化能力弱。本文提出IRPO，一种基于GRPO的确定性修复模型后训练框架。IRPO围绕数据构建与奖励建模两大核心：数据方面，选取预训练阶段表现最差的30%样本，提升准确率与训练效率；奖励方面，融合保真度导向与质量感知反馈，包含通用奖励（结构保真）、专家奖励（用视觉语言模型粗判视觉质量）和修复奖励（任务特异性低层线索）。在六项域内与五项域外（OOD）基准测试中，IRPO相较AdaIR基线，在域内任务提升0.93 dB，域外任务提升3.43 dB。代码已开源。
原文摘要 · Abstract (English)
Post-training has become effective for high-level generation, but its role in low-level vision remains underexplored. Existing image restoration methods often rely on fixed pixel-wise fitting to ground-truth images, which can lead to over-smoothing and weak generalization. We propose IRPO, a GRPO-based post-training framework for deterministic restoration models. IRPO is built around two axes: data formulation and reward modeling. For data formulation, we select the 30% underperforming samples from the pre-training stage, which improves both accuracy and training efficiency. For reward modeling, we combine fidelity-oriented and quality-aware feedback with three components: a General Reward for structural fidelity, an Expert Reward that uses a Vision-Language Model as a coarse visual-quality judge, and a Restoration Reward for task-specific low-level cues. Experiments on six in-domain and five out-of-domain (OOD) benchmarks show that IRPO improves the AdaIR baseline by 0.93 dB on in-domain tasks and 3.43 dB on OOD settings. Our code can be shown in https://github.com/HaoxuanXU1024/IRPO.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。