DELINE8K: A Synthetic Data Pipeline for the Semantic Segmentation of Historical Documents.
Taylor ArchibaldTony MartinezPublished in: CoRR (2024)
Keyphrases
- synthetic data
- semantic segmentation
- historical documents
- handwriting recognition
- conditional random fields
- superpixels
- word recognition
- document images
- scene classification
- weakly supervised
- object categories
- object class
- data sets
- pascal voc
- image understanding
- real world
- object recognition
- object classes
- object detection
- multiscale
- bayesian networks
- speech recognition
- image representation
- vision system
- real image data
- image set
- character recognition
- object segmentation
- markov random field
- information extraction
- probabilistic model
- viewpoint