A Comparative Study of Layout Based and OCRFree Models for Turkish Receipt Documents

Reading and understanding invoice and receipt documents accurately is a challenging process. Complex and costly procedures are required for a comprehensive understanding of the documentation. Generally, in these processes, Optical Character Recognition (OCR) techniques are used to read invoice and receipt documents. In recent years, with the development of transformer architecture, different methods have also been employed.In this study, the performances of the LayoutLMv3 model using OCR technique and the OCR-Free Donut models were tested. The tests were conducted using a dataset consisting of Turkish content invoice and receipt images. The created dataset was also used to test the models' performance in the Turkish language