Penerapan Algoritma Backpropagation Untuk Text Recognition Yang Ditranslate Ke Bahasa Daerah

Abstract

Every area of the community there is information, news, announcements or notifications through print media such as pictures, banners, posters and paper. However, there are many foreign languages (English) in the information which makes people not understand the language, because people only understand their own regional language (Sundanese) and the lack of media to help people understand the regional language (Sundanese) to a foreign language (English). Therefore, the author takes research on android-based text detection. Text detection or called Optical Character Recognition (OCR) is a system that can detect text into data files that can be processed in such a way. The method of collecting data or information in this study uses qualitative, the method for developing this system uses the waterfall method which has the advantage of being able to accelerate system development and using a backpropagation algorithm for text recognition. Backpropagation algorithm is an algorithm that can minimize errors, so as to make text detection more accurate. The results of the text detection accuracy are 97% and the error accuracy results are 3%, the text accuracy results are taken from a total of 14 samples, and 68 words. The results of text detection are affected by several cases such as the type of text font, the level of light in the image and the direction in which the photo or image is taken.

Keywords

• ml-kit, text recognition, translation, backpropagation, android