Subway Station Signage OCR Recognition Dataset

In the current transportation industry, the signage systems at subway stations often suffer from low visual recognition efficiency, especially in complex environments, leading to passengers getting lost and affecting their travel experience. Existing OCR recognition solutions largely rely on manual labeling, which is inefficient and has issues of inconsistency. This dataset aims to address the automatic recognition needs of subway station signage by providing rich annotated image data to enhance the application capabilities of OCR technology in this field. Data collection was carried out using high-definition camera equipment in various subway station environments, ensuring coverage of different signage styles and lighting conditions. For quality control, a method combining multiple rounds of annotation and expert review is adopted to ensure consistency and accuracy of labels. Data is stored in JPG and JSON formats, with a one-to-one correspondence between image data and annotation information, facilitating subsequent training and testing. The core advantage of this dataset is its high annotation accuracy, achieving over 95% consistency, while introducing new data augmentation techniques to improve model robustness, expected to increase recognition accuracy by 20%. Additionally, a systematic quality evaluation method ensures dataset integrity, addressing common issues of uneven annotation quality in the industry.