PENGENALAN TULISAN KARAKTER MANDARIN (HANZI) DENGAN MENGGUNAKAN METODE EFFICIENTDET

Abstract

In today's era of globalization and digitalization, the need to understand and recognize language differences is becoming increasingly important, especially with the increasing cross-cultural interactions. Mandarin, as the second international language, one of the languages ​​of the United Nations (UN), and one of the languages ​​with the largest number of native speakers in the world, has a complex writing system consisting of thousands of characters called Hanzi. Each of these characters not only represents a sound but also a specific meaning, making the recognition of Mandarin characters a challenge. In this study, a system was developed to recognize 200 Chinese characters using the EfficientDet method, an object detection model developed by researchers from Google Research. EfficientDet is known for its efficiency in detecting objects with high accuracy while maintaining fast processing speed. In conclusion, the system obtained overall mAP results of 58%, mAR 42.64%, IoU 49.18%, precision 75%, recall 64%, and F1-score 65%.