Table Structure Recognition of Historical Dongba Documents.

The analysis of table structures in historical documents has been a crucial area of research. Its objective is to identify the location of tables within the documents and segment each table cell. However, this process is often hindered by irregular structures and damage caused by inadequate preservation. Similarly, analyzing historical records written in the pictographic Dongba script of the Naxi people presents significant challenges. In this paper, we propose a novel method for extracting the table structure of historical Dongba documents. Our approach involves annotating the document image with specific patterns and applying an object detection model to precisely locate the regions of interest in the image. We then extract the table structure using a set of post-processing techniques. We evaluate the effectiveness of our proposed method by conducting experiments using various object detection models and providing a detailed analysis of our dataset in the Experiments section. Our results demonstrate that our approach is highly beneficial.