AesFormer: Transform Everyday Photos into Beautiful Memories
Researchers have introduced AesFormer, a novel two-stage framework designed to enhance the aesthetic quality of everyday photographs by addressing structural flaws. The system first plans editing actions based on aesthetic principles and then executes these structural edits. To support this work, a new benchmark called AesRecon, comprising over 9,000 image pairs, was created, and experiments demonstrate AesFormer's effectiveness in improving aesthetic photo reconstruction. AI
IMPACT This research introduces a new method for improving photo aesthetics, potentially impacting image editing tools and user-generated content.