TC-OCR: TableCraft OCR for Efficient Detection & Recognition of Table Structure & Content

TL;DR

The paper introduces TC-OCR, an end-to-end pipeline that unifies table detection, structure recognition, and content extraction by integrating DETR, CascadeTabNet, and PP OCR v2 to robustly interpret image-based tables across diverse layouts. The approach addresses limitations of prior methods by preserving table structure and enabling accurate data extraction, achieving an IOU of $0.96$ and OCR accuracy of $78\%$, with substantial improvements over the Table Transformer baseline. Experimental results on the TableBank dataset demonstrate faster inference and superior multimodal table understanding, highlighting the practicality for information retrieval and knowledge graph enrichment. The work underscores the potential of combining transformer-based detection with advanced segmentation and OCR in end-to-end table understanding, while acknowledging limitations with complex merged/nested tables and outlining avenues for multimodal extensions.

Abstract

The automatic recognition of tabular data in document images presents a significant challenge due to the diverse range of table styles and complex structures. Tables offer valuable content representation, enhancing the predictive capabilities of various systems such as search engines and Knowledge Graphs. Addressing the two main problems, namely table detection (TD) and table structure recognition (TSR), has traditionally been approached independently. In this research, we propose an end-to-end pipeline that integrates deep learning models, including DETR, CascadeTabNet, and PP OCR v2, to achieve comprehensive image-based table recognition. This integrated approach effectively handles diverse table styles, complex structures, and image distortions, resulting in improved accuracy and efficiency compared to existing methods like Table Transformers. Our system achieves simultaneous table detection (TD), table structure recognition (TSR), and table content recognition (TCR), preserving table structures and accurately extracting tabular data from document images. The integration of multiple models addresses the intricacies of table recognition, making our approach a promising solution for image-based table understanding, data extraction, and information retrieval applications. Our proposed approach achieves an IOU of 0.96 and an OCR Accuracy of 78%, showcasing a remarkable improvement of approximately 25% in the OCR Accuracy compared to the previous Table Transformer approach.

Figures (2)

• Figure 1: Our TC-OCR achieves simultaneous Table Detection (TD), Table Structure Recognition (TSR), and Table Content Recognition (TCR), preserving table structures and accurately extracting tabular data from document images.
• Figure 2: Architecture of the proposed Methodology, where we have incorporated three distinct models DETR for table detection, CascadeTabNet for table structure recognition, and PP OCRv2 for text detection and recognition