Optical Character Recognition – Machine Learning
An image-processing project that locates text in an image and converts it to a machine-readable label.
OCR (Computer Vision)
OCR is a project in the image processing field. It is based on an algorithm that analyses a given image and finds out where text is placed on it.
The algorithm first analyses the picture using convolutional neural networks (CNN) and passes the results — a number of features — to a recurrent neural network (RNN), which predicts a character for each feature. Connectionist temporal classification (CTC) then converts those predictions to the text label. It is also possible to use other types of classification algorithm, such as naive Bayes or a hidden Markov model.
Uses of the OCR algorithm include conversion of handwritten text to machine-readable text, extracting contact information, passport recognition in airports, aid for blind people, and a preprocessing step in more complex systems such as a license plate recognizer.
