Towards Khmer Scene Document Layout Detection

2026-02-28

AI Summary

This study addresses the challenge of layout analysis in Khmer scene documents, where scarce annotated data, complex backgrounds, and perspective distortions hinder the performance of existing Latin-based models in accurately detecting dense text regions. To tackle this issue, we present the first systematic investigation of Khmer document layout analysis, introducing the first publicly available dataset for Khmer scene document layout detection. We propose a YOLO-based baseline model that integrates synthetic scene image augmentation with oriented bounding boxes (OBBs) to effectively handle geometric distortions and the unique structural characteristics of Khmer script. The project releases the dataset, code, and trained models openly, significantly advancing layout recognition performance for Khmer documents and offering a comprehensive solution for document analysis in low-resource languages.

Technology Category

• Computer Vision: Scene Analysis & Understanding
• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

While document layout analysis for Latin scripts has advanced significantly, driven by the advent of large multimodal models (LMMs), progress for the Khmer language remains constrained because of the scarcity of annotated training data. This gap is particularly acute for scene documents, where perspective distortions and complex backgrounds challenge traditional methods. Given the structural complexities of Khmer script, such as diacritics and multi-layer character stacking, existing Latin-based layout analysis models fail to accurately delineate semantic layout units, particularly for dense text regions (e.g., list items). In this paper, we present the first comprehensive study on Khmer scene document layout detection. We contribute a novel framework comprising three key elements: (1) a robust training and benchmarking dataset specifically for Khmer scene layouts; (2) an open-source document augmentation tool capable of synthesizing realistic scene documents to scale training data; and (3) layout detection baselines utilizing YOLO-based architectures with oriented bounding boxes (OBB) to handle geometric distortions. To foster further research in the Khmer document analysis and recognition (DAR) community, we release our models, code, and datasets in this gated repository (in review).

Problem

• Khmer script
• scene document layout detection
• document layout analysis
• annotated training data scarcity
• geometric distortions

Innovation

• Khmer document layout
• scene document synthesis
• oriented bounding box
• data augmentation
• YOLO-based layout detection