AesFormer: 将日常照片转化为美丽的记忆
AesFormer: Transform Everyday Photos into Beautiful Memories
AI总结：
本文提出AesFormer框架，通过将审美规划与图像编辑解耦，改进照片的审美质量，同时保持主体身份和场景语义，构建了包含9071对严格对齐图像对的AesRecon基准数据集。
摘要：
在日常摄影中，吸引人的时刻往往受到结构缺陷（如构图、相机视角或姿势）的影响，而现有的修图和人像增强方法无法修复这些缺陷。我们提出将审美照片重建（APR）定义为通过结构重建来提高照片的审美质量，同时保持主体身份和场景语义。尽管最近的图像编辑模型使APR成为可能，但它们通常缺乏审美理解，导致编辑结果在语义上合理但审美上薄弱。为此，我们提出了AesFormer，一个两阶段框架，将审美规划与图像编辑解耦。在第一阶段，一个审美动作模型（AesThinker）分析输入沿七个渐进的摄影维度，并输出可执行的编辑动作；我们进一步应用GRPO-A来鼓励在多样化的动作计划上进行广泛探索，超越SFT。在第二阶段，一个动作条件编辑器（AesEditor）在这些动作的指导下执行结构编辑。为了支持APR，我们构建了一个基于视频的语料挖掘管道（VCMP）并构建了AesRecon，一个包含9,071对严格对齐（差，好）图像对的基准。实验表明，AesFormer显著提高了APR性能，并与Nano Banana Pro具有竞争力。代码可在https://github.com/PKU-ICST-MIPL/AesFormer_ICML2026获取。
英文摘要：
In everyday photography, aesthetically appealing moments are often captured with structural flaws (e.g., composition, camera viewpoint, or pose) that existing retouching and portrait enhancement methods cannot fix. We formulate Aesthetic Photo Reconstruction (APR) as improving a photo's aesthetic quality via structural reconstruction while preserving subject identity and scene semantics. Although recent advances in image editing models make APR feasible, they often lack aesthetic understanding, yielding edits that are semantically plausible yet aesthetically weak. To address this, we propose AesFormer, a two-stage framework that decouples aesthetic planning from image editing. In Stage 1, an aesthetic action model (AesThinker) analyzes the input along seven progressive photographic dimensions and outputs executable editing actions; we further apply GRPO-A to encourage broad exploration over diverse action plans beyond SFT. In Stage 2, an action-conditioned editor (AesEditor) performs structural edits guided by these actions. To support APR, we build a video-based corpus-mining pipeline (VCMP) and construct AesRecon, a benchmark of 9,071 strictly aligned (poor, good) image pairs. Experiments show that AesFormer substantially improves APR performance and is competitive with Nano Banana Pro. Code is available at https://github.com/PKU-ICST-MIPL/AesFormer_ICML2026.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
