IRPO: Boosting Image Restoration via Post-training GRPO
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911722269310976 |
|---|---|
| author | Xu, Haoxuan Liu, Yi Li, Tianfu Shen, Ruolin Jiang, Boyuan Peng, Jinlong Luo, Donghao Hu, Xiaobin Yan, Shuicheng Li, Haoang |
| author_facet | Xu, Haoxuan Liu, Yi Li, Tianfu Shen, Ruolin Jiang, Boyuan Peng, Jinlong Luo, Donghao Hu, Xiaobin Yan, Shuicheng Li, Haoang |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_00814 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | IRPO: Boosting Image Restoration via Post-training GRPO Xu, Haoxuan Liu, Yi Li, Tianfu Shen, Ruolin Jiang, Boyuan Peng, Jinlong Luo, Donghao Hu, Xiaobin Yan, Shuicheng Li, Haoang Computer Vision and Pattern Recognition 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. |
| title | IRPO: Boosting Image Restoration via Post-training GRPO |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2512.00814 |