A reminder: tomorrow at 14 in room 107 (via di S Marta 3 Firenze) Zarah will give a talk on "Document Analysis by Deep Learning"
Abstract
The technology of document analysis and recognition (DAR) is to analyze
the structure and textual contents of document images and handwriting.
It faces numerous application needs such as digitization of books and
forms, pen-based text input, information extraction from Web document
images. It has been under study as a field of pattern recognition since
1960s. In recent years, the introduction of deep learning to DAR has led
to significant improvement of performance in many branches, particularly
in the cases when large sets of labeled data are available for
supervised learning, such as handwritten character and text recognition.
Among the most successful deep learning models are the convolutional
neural network (CNN) and the recurrent neural network with long
short-term memory (LSTM). The application of deep learning is now
extended to scene text detection and recognition, document image
segmentation and layout analysis, writer identification, document
retrieval, and so on.