Recognition Systems
English IAM-DB HMM-Based Recognizer with 20k Word Vocabulary and Bigram LM
An HMM-based recognition system that models each character with an individual number of states. Twelve Gaussians are used to model the state transition probabilities. The recognizer is lexicon-driven. The vocabulary consists of the 20,000 most frequent words that occur in the LOB, the Brown, and the Wellington corpus. The recognition is supported by a statistical bigram langauge model. This recognizer is similar to the recognizer used as reference system in: R. Bertolami, S. Uchida, M. Zimmermann, and H. Bunke. Non-uniform slant correction for handwritten text line recognition. In Proc. 9th International Conference on Document Analysis and Recognition, Curitiba, Brazil, pages 18–22, 2007.
German Text Recognizer (Optimized for the Handwriting of Gerhard Meier)
An HMM-based recognition system that models the handwriting of the Swiss writer Gerhard Meier. The vocabulary contains about 2000 German words. The recognition is supported by a statistical bigram langauge model.
English IAM-DB Neural Network Recognizer with 20k Word Vocabulary
A recognition system based on a recurrent neural network that includes a bidirectional Long Short-Term Memory (BLSTM) architecture and a 20,000 word vocabulary. This recognizer is based on the recognizer used in: A. Graves, M. Liwicki, S. Fernández, R. Bertolami, H. Bunke, and J. Schmidhuber. A novel connectionist system for improved unconstrained handwriting recognition. accepted for publication at IEEE Transactions on Pattern Analysis and Machine Intelligence.
English IAM-DB Neural Network Recognizer with 20k Word Vocabulary and Bigram LM
A recognition system based on a recurrent neural network that includes a bidirectional Long Short-Term Memory (BLSTM) architecture. The recognition is performed supported by a statistical bigram language model based on a 20,000 word vocabulary. This recognizer is based on the recognizer used in: A. Graves, M. Liwicki, S. Fernández, R. Bertolami, H. Bunke, and J. Schmidhuber. A novel connectionist system for improved unconstrained handwriting recognition. accepted for publication at IEEE Transactions on Pattern Analysis and Machine Intelligence.
