back-kh/SADA-Ancient-Palm-Leaf-Manuscripts-Recognitions

[PRL 2025, APSIPA 2022] Syllable Analysis Data Augmentation (SADA), This project introduces a glyph dictionary and grammar-aware augmentation strategy designed to enhance Khmer palm leaf manuscript recognition. By modeling the language's grammatical structure, we support more robust OCR performance in low-resource settings. # Syllable Analysis Data Augmentation (SADA) ### This repository supports two research projects focused on text recognition for ancient Southeast Asian scripts. ### (**One of these approaches won First Place in Text Recognition Tasks at ICFHR 2018 for Palm Leaf Manuscript Analysis.** ) ### Additionally, the projects incorporate syllable-based data augmentation. These efforts aim to enhance recognition systems for palm leaf manuscripts by integrating linguistic grammar into visual recognition models. --- ## 📄 Papers - **Paper 1:** *Khmer Syllable Analysis Data Augmentation (KH-SADA)* - **Paper 2:** *Multi-Low Resource Languages Syllable Analysis Data Augmentation (PALM-SADA) --- ## Paper 1: KH-SADA – Khmer Syllable Analysis Data Augmentation ### 🧠 Introduction This project introduces a glyph dictionary and grammar-aware augmentation strategy designed to enhance Khmer palm leaf manuscript recognition. By modeling the language's grammatical structure, we support more robust OCR performance in low-resource settings. Figure 1: Complexity of a Khmer word with stacked consonants and multiple diacritics. ### 🗂️ Background Inspired by publicly available datasets from the ICFHR 2018 competition, we built this dictionary to capture essential grammar forms of Khmer syllables. This linguistic grounding is critical for generating valid word formations and improving recognition accuracy. Figure 2: Complexity of Khmer grammar structure with syllabic combinations and contextual glyph roles. ### 📦 Repository Contents - **Khmer Grammar Structures** Core grammar forms (CV, CVC, etc.) used in valid word formations. - **Glyph Classes & Types** Categorized glyphs (e.g., base consonants, subscripts, vowels) to facilitate structured augmentation. Figure 3: Component breakdown of Khmer writing forms, including consonants and dependent vowels. - **Ground Truth Training Data** Labeled glyph image dataset used for training and evaluation. - **Latin Transliterated Di …