OCR Datasets

This dataset covers 21 languages, with 20,000 to 25,000 images per language. The data includes natural scenes, document photography scenes, and electronic scenes. The data diversity includes various data types, multiple shooting angles, and multiple languages. In terms of annotation, quadrilateral or polygonal at the row (column) level and content transcription at the row (column) level are adopted. This dataset can be use for multilingual optical character recognition (OCR) and text detection tasks.

This dataset contains 1,000 Japanese invoice images, it includes 500 images with basic virtual editing and 500 images with professional editing. Data diversity includes different invoice contents, different editing types, and multiple invoice formats. The company name, address, name, fax number, phone number and other sensitive information on the invoice have been virtually edited and are not real information. The data can be used for tasks such as invoice detection, recognition, and end-to-end OCR tasks.

426,687 high-resolution images featuring multilingual Optical Character Recognition (OCR) data across both natural scenes and various document types. This dataset spans 20 languages, including Traditional Chinese, Simplified Chinese, Japanese, Korean, Thai, Vietnamese, Indonesian, Malay, Polish, and more. The data covers a wide range of real-world conditions—natural scenes, printed documents, handwritten notes, signs, and posters—captured from multiple countries and environments, with varied backgrounds, lighting conditions, and camera angles. All images are annotated for OCR tasks, making this dataset highly suitable for training deep learning models for text detection, recognition, and layout analysis in multi-language scenarios. The dataset complies with global data protection standards (GDPR, CCPA, PIPL), and is validated by leading AI enterprises for commercial and research applications.

29,954 Images - OCR Collection Data of Southeast Asian Languages, including Khmer (Cambodia), Lao and Burmese. The diversity of collection includes multiple languages, multiple collection types, multiple shooting angles. This set of data can be used for Southeast Asian language OCR tasks.

This dataset contains 222,289 images of Chinese text in natural scenes. The collecting scenes of this dataset include indoor and outdoor scenes. The data diversity includes multiple scenes, and multiple shooting angles. For annotation, we have annotated in line-level, word-level, and character-level and well content matched text transcription included. The dataset can be used for OCR tasks in natural scenes.

This dataset contains 9,497 images of 10 types of forms. Rectangular bounding boxes were adopted to annotate forms. The dataset can be used for tasks such as forms detection.

This dataset contains 17,561 images of primary school mathematics papers. The data feature a pure color background. The data covers multiple question types, multiple types of test papers (math workbooks, test papers, competition test papers, etc.) and multiple grade levels. The dataset can be used for tasks such as intelligent scoring and homework guidance for primary school students.

This dataset contains 4,995 Vietnamese OCR images with annotations and text transcriptions. The data includes 258 natural scene images, 2,553 Internet images, and 2,184 document images. For line-level content annotation, quadrilateral bounding box annotations and text transcriptions are provided. For column-level content annotation, column-level quadrilateral bounding box annotation and text transcription are provided. The data can be used for tasks such as Vietnamese recognition in multiple scenes.

This dataset contains 104,320 images of Korean and Hindi text in natural scenes. The collecting scenes include packaging, posters, tickets, reminders, menus, building signs, and similar real-world environments. The dataset features diverse scenes, multiple shooting angles and multiple lighting conditions. For annotation, line-level polygon bounding box (or tetragon or rectangular bounding boxes) annotation, transcription and text attributes (language type) for the texts; vertical-level polygon bounding box (or tetragon or rectangular bounding boxes) annotation, transcription and text attributes (language type) for the text. The dataset can be used for Korean and Hindi OCR tasks in natural scenes.