dr. István Marosi

Name: dr. István Marosi
Affiliation: Tungsten-Recognita Corp
Primary research interest: Image processing and recognition
Title of the lecture: Deep Learning Techniques in OmniPage OCR application
Keywords: OCR, machine print, Convolutional NN, Recurrent NN
Summary: Industrial strength Optical Character Recognition applications usually apply traditional image processing and neural network approaches. The current boom in AI together with advanced CPUs make it possible to implement Deep Learning algorithms in these general purpose applications, too. Some examples will be shown on how we use CNN (convolution) and RNN (recurrent) neural network based algorithms in image pre-processing and character recognition in our OmniPage application.