Multi-Branch Deep Learning for Non-Normative Table Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing table detection methods struggle with non-normative tables, such as those without clear boundary lines, multi-column layouts, and complex structures, particularly in medical detailed bills, leading to poor detection accuracy and usability.
Innovation Solution
A parallel multi-branch table detection method using a convolutional neural network to learn pixel-level category information, predicting full image, table columns, and table header rows, which strengthens detection accuracy by considering prior table structure information and training a deep learning model with sample images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional table detection methods are used, then the detection process is simple, but the detection accuracy is poor for non-normative tables
Solution Approach 1:
The detection model is divided into three independent detection branches: full table detection branch, column detection branch, and header detection branch. Each branch focuses on detecting specific table components, allowing the system to handle complex non-normative tables by processing different structural elements separately and then integrating the results.
Solution Approach 2:
The patent introduces multi-dimensional detection by not only detecting the overall table boundary but also separately detecting column positions and header regions. This multi-dimensional approach adds spatial dimensionality to the detection process, enabling accurate identification of tables with complex layouts, merged cells, and irregular structures that traditional single-dimension methods cannot handle.
2Productivity
If a pre-trained deep learning model is used, then the detection speed is fast, but the model requires significant training data and computational resources
Solution Approach 1:
The system employs a pre-trained deep learning model that has been previously trained on a large dataset of table images. This preliminary training action allows the model to learn general table structures, patterns, and features in advance, so that during actual detection operations, the model can quickly process new images without requiring additional training data or computational resources at runtime.
3Measurement precision
If multi-branch detection is implemented, then the detection accuracy for complex tables is improved, but the computational complexity increases
Solution Approach 1:
By segmenting the detection task into three specialized branches (full table, column, header), each branch processes only the specific component it is designed for, reducing the computational burden on each individual branch compared to a single comprehensive detection model. This segmentation allows parallel processing and optimizes resource utilization.
Solution Approach 2:
The three detection branches share a common feature extraction backbone, allowing the system to extract features once and then reuse them across multiple detection tasks. This multi-functionality approach reduces redundant computation and energy consumption while maintaining the ability to detect various table components accurately.
Data Source
AI summary
A method and apparatus for detecting a table. The method includes: acquiring a to-be-processed image; inputting the to-be-processed image into a pre-trained deep learning model, and outputting a full table detection branch result, a column detection branch result and a header detection branch result through the deep learning model; where the full table detection branch result represents a detection result for a full table in the to-be-processed image, the column detection branch result represents a detection result for a column in the table in the to-be-processed image, and the header detection branch result represents a detection result for a header in the to-be-processed image; and obtaining a detection result of the table in the to-be-processed image, based on the full table detection branch result, the column detection branch result and the header detection branch result.