Improving OCR Accuracy on Early Printed Books by utilizing Cross Fold Training and Voting.
Christian ReulUwe SpringmannChristoph WickFrank PuppePublished in: CoRR (2017)
Keyphrases
- optical character recognition
- high accuracy
- computational cost
- prediction accuracy
- character recognition
- scanned documents
- preprocessing
- error correction
- computational complexity
- training algorithm
- test set
- post processing
- training set
- data sets
- high precision
- supervised learning
- training phase
- printed documents
- voting method
- training speed
- recognition errors
- training and testing data