Historical Document Image Segmentation with LDA-Initialized Deep Neural Networks.
Michele AlbertiMathias SeuretVinaychandran PondenkandathRolf IngoldMarcus LiwickiPublished in: HIP@ICDAR (2017)
Keyphrases
- image segmentation
- neural network
- topic discovery
- latent dirichlet allocation
- latent dirichlet
- linear discriminant analysis
- latent topics
- topic models
- pattern recognition
- latent semantic analysis
- document retrieval
- artificial neural networks
- generative model
- discriminant analysis
- web documents
- lda model
- multiscale
- face recognition
- dimensionality reduction
- document classification
- retrieval systems
- feature extraction
- markov random field
- computer vision
- text documents
- information retrieval systems
- keywords
- fuzzy logic
- document collections
- segmentation method
- document images
- segmentation algorithm
- graph cuts
- genetic algorithm
- region growing
- topic modeling
- document analysis
- back propagation
- historical documents
- neural network model
- dimension reduction
- relevant documents
- deformable models
- energy function
- active contours
- principal component analysis
- image processing
- information retrieval
- historical data
- query expansion
- level set