HTR for Greek Historical Handwritten Documents

Abstract

Offline handwritten text recognition (HTR) for historical documents aims for effective transcription by addressing challenges that originate from the low quality of manuscripts under study as well as from several particularities which are related to the historical period of writing. In this paper, the challenge in HTR is related to a focused goal of the transcription of Greek historical manuscripts that contain several particularities. To this end, in this paper, a convolutional recurrent neural network architecture is proposed that comprises octave convolution and recurrent units which use effective gated mechanisms. The proposed architecture has been evaluated on three newly created collections from Greek historical handwritten documents that will be made publicly available for research purposes as well as on standard datasets like IAM and RIMES. For evaluation we perform a concise study which shows that compared to state of the art architectures, the proposed one deals effectively with the challenging Greek historical manuscripts.

Keywords: convolutional neural networks; document image dataset; gated recurrent unit; handwritten text recognition; recurrent neural networks.

Figure 1

Schematic diagram of the proposed architecture, consisting of (a) the CNN stage and (b) the recurrent stage.

Figure 2

An example text line image (a) before and (b) after the preprocessing stage.

Figure 3

Feature maps produced by each layer of the Octave-CNN model.

Figure 4

An illustration of the gated recurrent unit (GRU) [8].

Figure 5

Example document image from the collection $$\chi \varphi$$53.

Figure 6

Example document image from the collection $$\chi \varphi$$79.

Figure 7

Example document image from the collection $$\chi \varphi$$114.

Figure 8

Floating characters appearing at word endings. The floating portion of the word is represented by a rectangle, while the rest of the word is underlined.

Figure 9

‘Minuscule’ writing example. Key locations in the text line that correspond to this particularity are underlined.

Figure 10

Polytonic orthography example.

Figure 11

An example of a correctly predicted text line image along with the corresponding (a) groud-truth and (b) predicted texts.

Figure 12

An example of a problematic text line image along with the corresponding (a) groud-truth and (b) predicted texts. The errors concern diacritics (circle), spacing (red line) and abbreviations (square).