|
|
Semiautomatic Collection of Large Database of Handwritten Letters
Štěpánek, Ivo ; Juránek, Roman (referee) ; Herout, Adam (advisor)
This thesis deals with creation of database of handwritten characters further usable for handwritten character recognition. The issue of systems for unconstrained handwritten text recognition and datasets usable for it is discussed. Practical part of the thesis aims at preprocessing of the input document and line and word segmentation followed by extraction of isolated characters. These phases can be done entirely automatically, though user input to correct output of automatic processing is supposed. Furthermore the practical part is devoted to annotation of obtained character and to generation of XML document containing annotation and position of single characters from the input texxt. The created system is finally evaluated with emphasis on GUI and automatic segmentation succes rate.
|
|
|
Convolutional Networks for Handwriting Recognition
Sladký, Jan ; Kišš, Martin (referee) ; Hradiš, Michal (advisor)
This thesis deals with handwriting recognition using convolutional neural networks. From the current methods, a network model was chosen to consist of convolutional and recurrent neural networks with the Connectist Temporal Classification. The Vertical Attention Module, which selects the relevant information in each column corresponding to the text in the figure was subsequently implemented in such a model. Then, this module was compared with other possibilities of vertical aggregation between convolutional and recurrent networks. The experiments took place on a data set containing over 80,000 lines of text from Czech letters from the 20th century. The results show that the Vertical Attention Module almost always achieves the best results on all used types of convolution networks. The resulting network achieved the best result with 8,9% of the character error rate. The contribution of this work is a neural network with a newly introduced element that can recognize lines of text.
|
|
|
Convolutional Networks for Handwriting Recognition
Sladký, Jan ; Kišš, Martin (referee) ; Hradiš, Michal (advisor)
This thesis deals with handwriting recognition using convolutional neural networks. From the current methods, a network model was chosen to consist of convolutional and recurrent neural networks with the Connectist Temporal Classification. The Vertical Attention Module, which selects the relevant information in each column corresponding to the text in the figure was subsequently implemented in such a model. Then, this module was compared with other possibilities of vertical aggregation between convolutional and recurrent networks. The experiments took place on a data set containing over 80,000 lines of text from Czech letters from the 20th century. The results show that the Vertical Attention Module almost always achieves the best results on all used types of convolution networks. The resulting network achieved the best result with 8,9% of the character error rate. The contribution of this work is a neural network with a newly introduced element that can recognize lines of text.
|
|
|
Semiautomatic Collection of Large Database of Handwritten Letters
Štěpánek, Ivo ; Juránek, Roman (referee) ; Herout, Adam (advisor)
This thesis deals with creation of database of handwritten characters further usable for handwritten character recognition. The issue of systems for unconstrained handwritten text recognition and datasets usable for it is discussed. Practical part of the thesis aims at preprocessing of the input document and line and word segmentation followed by extraction of isolated characters. These phases can be done entirely automatically, though user input to correct output of automatic processing is supposed. Furthermore the practical part is devoted to annotation of obtained character and to generation of XML document containing annotation and position of single characters from the input texxt. The created system is finally evaluated with emphasis on GUI and automatic segmentation succes rate.
|
guest ::
