LayoutDIT: Layout-Aware End-to-End Document Image Translation with Multi-Step Conductive Decoder.
Zhiyang ZhangYaping ZhangYupu LiangLu XiangYang ZhaoYu ZhouChengqing ZongPublished in: EMNLP (Findings) (2023)
Keyphrases
- end to end
- multi step
- document images
- page layout
- rate allocation
- document layout
- document image analysis
- word level
- document analysis
- optical character recognition
- knn
- machine translation
- k nearest neighbor
- congestion control
- low complexity
- error concealment
- digital libraries
- structure extraction
- objective function
- scalable video
- scanned documents
- image sequences