Learning From Documents in the Wild to Improve Document Unwarping
Event Type
Technical Paper
Interest Areas
Research & Education
Presentation Types
In Person
Registration Categories
Full Conference Supporter
Full Conference
Exhibitor Additional Full Conference
Exhibitor Full Conference
This session WILL NOT be recorded.
TimeWednesday, 10 August 20229:50am - 9:55am PDT
LocationEast Building, Ballroom C
DescriptionDocument image unwarping is important for document digitization and analysis. In this work, we present the state-of-the-art, data-driven document unwarping approach, PaperEdge. Unlike synthetic image-based prior methods, PaperEdge can incorporate real-world images in training. PaperEdge significantly improves the generalizability and OCR performance on the unwarped images for different document types.