Layout-Aware Single-Image Document Flattening
DescriptionOur paper presents a novel framework for layout-aware single-image document flattening. It combines layout-aware segmentation, local and global UV map prediction, and an merging algorithm. Additionally, we introduce a new synthetic dataset containing meshes generated by a physics-based simulation, effectively supplementing existing datasets in terms of geometric deformation variety.
Authors
Event Type
Technical Papers
TimeTuesday, 3 December 20249:00am - 12:00pm JST
LocationHall C, C Block, Level 4
