Pharmacy Prescription Text Extraction Dataset

In the field of healthcare, digitizing pharmacy prescriptions is a crucial direction to improve medication management efficiency and reduce human errors. However, current solutions have significant limitations in text extraction accuracy and handwritten text recognition capabilities, affecting the practicality of electronic prescription systems. The construction of this dataset aims to address these issues by providing high-quality prescription image data to help improve the accuracy and reliability of text recognition systems. The dataset includes prescription images collected in various environments, involving different prescription formats and handwriting styles. High-definition scanners and professional photographic equipment were used during data collection to ensure image clarity. Quality control includes multiple rounds of manual annotation, cross-validation, and expert review, with the annotation team consisting of 20 professionals with pharmaceutical backgrounds. Data preprocessing includes image enhancement, noise reduction, and text area detection, finally stored in JPG format and organized by prescription type and date. The core advantage is that the dataset has an annotation accuracy of over 98% with highly consistent labeling, comprehensively covering both handwritten and printed text. Technological innovations include unique text differentiation and enhancement methods, increasing recognition accuracy by 15%. This dataset solves the problem of existing systems failing to accurately parse complex prescriptions, significantly improving the automation level and work efficiency of electronic pharmacies. Compared to similar datasets, our data quality is higher, with scarcity reflected in comprehensive coverage of multiple fonts and handwriting practices, offering good extensibility and versatility, suitable for the development and optimization of various pharmacy information systems.