Document element detection method, apparatus, device, and medium
By dual-training the detection model and utilizing the semantic features and labeled region differences of sample document images, the accuracy of document element detection was improved, thus enhancing the text recognition effect.
CN116363653BActive Publication Date: 2026-07-24IFLYTEK CO LTD
View PDF 5 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- IFLYTEK CO LTD
- Filing Date
- 2023-02-17
- Publication Date
- 2026-07-24
AI Technical Summary
Technical Problem
The accuracy of document element detection in existing technologies is insufficient, which affects the text recognition effect.
Method used
The detection model is trained by first training based on the differences between semantic features and image features of elements in sample document images, and second training based on the differences between labeled regions and predicted regions, thereby improving the accuracy of the detection model.
Benefits of technology
It improves the accuracy of document element detection and enhances the effectiveness of text detection.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116363653B_ABST
Abstract
A document element detection method, device, equipment and medium are disclosed. The detection method comprises: acquiring a to-be-detected document image; and detecting the to-be-detected document image by using a detection model to obtain a target region of a document element in the to-be-detected document image. The detection model is obtained through at least one of first training and second training. In the first training, the training is performed based on at least a difference between an element semantic feature and an element image feature of a sample document element in a sample document image. In the second training, the training is performed based on a difference between a labeled region and a predicted region of a sample document element in a sample document image. The predicted region is predicted based on at least a sample semantic feature and a sample image feature of the sample document image. In this way, the accuracy of document element detection can be improved.
Need to check novelty before this filing date? Find Prior Art