Mohammadian Mehrdad ; Maleki Neda ; Olsson Tobias ; Ahlgren Fredrik

Abstract

What happens if we encounter a suitable font for our design work but do notknow its name? Visual Font Recognition (VFR) systems are used to identify thefont typeface in an image. These systems can assist graphic designers inidentifying fonts used in images. A VFR system also aids in improving the speedand accuracy of Optical Character Recognition (OCR) systems. In this paper, weintroduce the first publicly available datasets in the field of Persian fontrecognition and employ Convolutional Neural Networks (CNN) to address thisproblem. The results show that the proposed pipeline obtained 78.0% top-1accuracy on our new datasets, 89.1% on the IDPL-PFOD dataset, and 94.5% on theKAFD dataset. Furthermore, the average time spent in the entire pipeline forone sample of our proposed datasets is 0.54 and 0.017 seconds for CPU and GPU,respectively. We conclude that CNN methods can be used to recognize Persianfonts without the need for additional pre-processing steps such as featureextraction, binarization, normalization, etc.