ConRTF: Edge-Constrained Boundary Distribution Refinement for Realtime TransFormer Table Structure Recognition

2026-07-01

AI Summary

This work addresses the challenge of inaccurate row/column boundary localization in table structure recognition, which often leads to erroneous cell assignments. Existing methods typically overlook the inherent geometric asymmetry between rows and columns. To remedy this, the authors propose an Edge-constrained Fine-grained Localization (EFL) loss that incorporates geometric priors during training—emphasizing horizontal boundaries for rows and vertical boundaries for columns—and integrates a Distribution-aware Boundary Refinement module (D-FINE) to enhance localization accuracy without increasing inference overhead. This approach is the first to explicitly model row-column asymmetry as part of the training objective, enabling efficient, structure-aware boundary optimization. Evaluated within a Transformer-based real-time detection framework, the method outperforms RT-DETRv2 and YOLOv10–11 on PubTables-1M and two private datasets, achieving up to a 1.6-point gain in GriTS scores while maintaining robust performance with only 2k–3k annotated samples.

Abstract

Table Structure Recognition (TSR) aims to recover the row and column layout of tables from document images, a key step in document understanding pipelines. Accurate TSR depends on precise boundary localization: small errors in row or column boundaries can propagate into incorrect cell assignments and structural inconsistencies. Yet detection-based approaches treat table elements as generic objects, ignoring a fundamental property of table layout: rows and columns play structurally distinct roles and their boundaries carry unequal importance. We propose an Edge-constrained Fine-grained Localization loss (EFL) that formalizes this structural asymmetry by encoding table-specific geometric priors into the training objective: row-like elements are supervised with emphasis on their horizontal boundaries, while column-like elements prioritize vertical boundaries. Implemented within a real-time detector with distribution-based boundary refinement (D-FINE), EFL operates during training only and guides boundary refinement toward structurally meaningful adjustments with no change to the inference pipeline. The proposed approach, ConRTF, is also data-efficient, maintaining robust accuracy with as few as 2k--3k annotated tables. Experiments on PubTables-1M and two private datasets show consistent improvements over the optimized baseline and several real-time detectors including RT-DETRv2 and YOLOv10-11, with gains of up to +1.6 GriTS points at equal inference speed.

Problem

• Table Structure Recognition
• Boundary Localization
• Structural Asymmetry
• Row-Column Layout
• Document Understanding

Innovation

• Edge-constrained Loss
• Table Structure Recognition
• Boundary Refinement
• Structural Asymmetry
• Real-time Detection