Recognition of Oracle Bone Inscriptions by using Two Deep Learning Models

Oracle bone inscriptions (OBIs) contain some of the oldest characters in the world and were used in China about 3000 years ago. As an ancient form of literature, OBIs store a lot of information that can help us understand world history, character evaluations, and more. However, as OBIs were only discovered about 120 years ago, few studies have described them, and the aging process has made the inscriptions less legible. Hence, automatic character detection and recognition have become important issues. This paper aims to design an online OBI recognition system for helping cultural heritage preservation and organization. We have evaluated two deep learning models for OBI recognition and designed an application programming interface (API) that can be accessed online for the recognition. In the first stage, You Only Look Once (YOLO) is applied for detecting and recognizing OBIs. However, since not all of the OBIs can be detected correctly by YOLO in the first stage, we then utilize MobileNet to recognize the undetected OBIs by manually cropping the undetected OBIs in the image. MobileNet is selected for the second stage of recognition as our evaluation of ten state-of-the-art deep learning models showed that it is the best network for OBI recognition due to its superior performance in terms of accuracy, loss and time consumption. We have installed our system on a server as an API and opened it for OBI detection and recognition.