CNN-Bidirectional LSTM Based Optical Character Recognition of Sanskrit Manuscripts : A Comprehensive Systematic Literature Review

Kataria, Bhavesh; Jethva, Harikrishna B.

doi:10.32628/cseit2064126

Cited by 161 publications

(1 citation statement)

References 63 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…YOLOv4 has been used to train a CNN model for recognizing printed Arabic characters, achieving an accuracy of 82.4% [10]. Other researchers explored methods for developing an OCR system for Sanskrit Manuscripts, using conventional feature extraction, heuristic methods, and machine learning approaches such as CNN, LSTM, or Bidirectional LSTM [11]. An OCR system for Javanese Script was created using Projection Profile Segmentation and Nearest Centroid Classifier, achieving a 93.88% success rate for line segmentation and 73.59% success rate for character segmentation, with a classification accuracy of 60.6% [12].…”

Section: Introductionmentioning

confidence: 99%

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

Prasetiadi,

Saputra,

Kresna

et al. 2023

join

View full text Add to dashboard Cite

Indonesia boasts a diverse range of indigenous scripts, called Nusantara scripts, which encompass Bali, Batak, Bugis, Javanese, Kawi, Kerinci, Lampung, Pallava, Rejang, and Sundanese scripts. However, prevailing character detection techniques predominantly cater to Latin or Chinese scripts. In an extension of our prior work, which concentrated on the classification of script types and character recognition within Nusantara script systems, this study advances our research by integrating object detection techniques, employing the YOLOv5 model, and enhancing performance through the incorporation of the U-Net model to facilitate the pinpointing of fundamental Nusantara script's character locations within input document images. Subsequently, our investigation delves into rearranging these character positions in alignment with the distinctive styles of Nusantara scripts. Experimental results reveal YOLOv5's performance, yielding a loss rate of approximately 0.05 in character location detection. Concurrently, the U-Net model exhibits an accuracy ranging from 75% to 90% for predicting character regions. While YOLOv5 may not achieve flawless detection of all Nusantara scripts, integrating the U-Net model significantly enhances the detection rate by 1.2%.

show abstract

Section: Introductionmentioning

confidence: 99%