LONTAR_DETC: Dense and High Variance Balinese Character Detection Method in Lontar Manuscripts

Abstract

This paper proposed LONTAR_DETC, a method to detect handwritten Balinese characters in Lontar manuscripts. LONTAR_DETC is a deep learning architecture based on YOLO. The detection of Balinese characters in Lontar manuscripts is challenging due to the characteristics of Balinese characters in Lontar manuscripts. Balinese characters in Lontar manuscripts are dense, overlapping, have high variance, contain noise, and classes of these characters are imbalanced. The proposed method consists of three steps, namely data generation, Lontar manuscript annotation, and Balinese character detection. The first step is data generation, in which synthetic images of original Lontar manuscript images are generated with enhanced image quality. The second step is data annotation to build a new Lontar manuscript dataset. As a result, we also propose the Handwritten Balinese Character of Lontar manuscript (HBCL_DETC) dataset, a novel Balinese character in Lontar manuscripts dataset. HBCL_DETC contains 600 images that consists of more than 100,000 Balinese characters annotated by experts. Finally, the third step is training the YOLOv4 detection model using the HBCL_DETC dataset. We created this dataset specifically for the task of detecting Balinese characters in Lontar manuscripts. To evaluate the reliability of the dataset, we experimented with three scenarios. In the first scenario, the detection model was trained using original images of Lontar manuscripts, in the second scenario the detection model was trained with the addition of augmented grayscale images, and in the third scenario the detection model was trained using HBCL_DETC. Based on the experimental results, LONTAR_DETC can detect Balinese characters with high detection rate with mAP of 99.55%.

Augmentasi Data pada Prasasti Logam untuk Deteksi Aksara Kawi

“…Berkaitan dengan solusi pembacaan aksara pada prasasti, beberapa penelitian telah menerapkan deep learning dengan arsitektur CNN (Convolutional Neural Network) untuk melakukan deteksi dan pengenalan aksara pada prasasti seperti pada [10] dan [11]. Kedua penelitian ini menghasilkan performa model dengan akurasi pendeteksian yang tinggi.…”

“…Kedua penelitian ini menghasilkan performa model dengan akurasi pendeteksian yang tinggi. Penelitian pertama dilakukan oleh Suciati, dkk [10] menerapkan YOLOv4 pada dataset Aksara Bali yang terdapat pada prasasti bermedia lontar untuk melakukan pendeteksian aksara. Penelitian ini menerapkan beberapa skenario yang melibatkan proses augmentasi data.…”

DeepLontar dataset for handwritten Balinese character detection and syllable recognition on Lontar manuscript

“…The resulting performance is only below 50% due to the use of isolated character images, which do not label every character in the Balinese Lontar manuscript. Other studies related to Balinese characters have been carried out, starting with Balinese character segmentation 2 , Balinese character recognition 3 , Balinese character augmentation in increasing data variation 4 , and Balinese character detection based on deep learning 5 . In the case of ancient Chinese documents, two main datasets were proposed.…”

“…In the third dataset, another 200 augmented images (produced by adaptive gaussian thresholding technique) were added into the train data. In those three dataset, the YOLOv4 model produces a detection performance with mean average precision (mAP) of up to 99.55% with precision, recall, and F1-score are 99%, 100%, and 99%, respectively 5 . DeepLontar consists of 55 Balinese character classes.…”

YOLOv5 and U-Net-based Character Detection for Nusantara Script

“…The Faster Region-based Convolutional Neural Network (Faster R-CNN) has also been employed in Javanese script detection, achieving accuracy ranging from 41.67% to 96.31% [13]. YOLOv4 has demonstrated an impressive detection rate of nearly 99.55% for Balinese scripts on traditional lontar manuscripts [14]. Despite various versions of YOLO, YOLOv5 outperforms other versions in challenging environments like underwater scenarios [15].…”