OCR Datasets

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.

128,900 Images-Multiple scenes OCR Data of Seven Languages. This data have seven languages, including Arabic, French, German, Hindi, Italian, Portuguese, and Spanish. There are two scenes of this data like blur and nature, and two special shooting styles like handheld shooting and moire. The annotation of this data includes polygonal bounding box labeling of text (with a precision between rectangular bounding box labeling and image segmentation labeling) and transcription. This data can be used for OCR tasks.

1,586,458 Sets Document OCR and Structured Analysis Data, Including Chinese Textbooks, Chinese E-books, Chinese Teaching Reference Books, etc . The annotated files include OCR annotations and structured analysis.

9,401 Images-English Document OCR Data. The language and text contents of this data are English and play script, book, exam paper, etc. The annotation of this data includes polygonal bounding box labeling of text (with a precision between rectangular bounding box labeling and image segmentation labeling) and transcription. This data can be used for English document OCR tasks.

30,276 Images-English Handwriting OCR Data. The language and writing style of this data are English and horizontal left-to-right writing, including different handwriting styles, different text colors (black, blue, red). There are two text mediums of this data like A4 paper and lined paper. The annotation of this data includes polygonal bounding box labeling of text (with a precision between rectangular bounding box labeling and image segmentation labeling) and transcription. This data can be used for English handwriting OCR tasks.

This dataset contains 5,156 images of handwritten mathematical formulas collected under diverse writing conditions, including A4 paper, grid paper, lined paper, whiteboards, and other writing surfaces. The dataset covers various mathematical expressions, handwriting styles, paper types, and capture conditions. Images were collected from multiple viewpoints, including upward-facing and eye-level camera angles. The dataset is designed for Math OCR, handwritten formula recognition, mathematical expression recognition, and AI model training for education and document intelligence applications.

This dataset contains 8,604 images collected from diverse real-world Arabic text scenes, covering various environments, shooting angles, and natural conditions. Each text instance is annotated with precise quadrilateral or polygon bounding boxes along with transcription labels, enabling accurate text localization and recognition. This data can be used for Arabic OCR applications and multilingual computer vision model training.

This dataset contains 57,645 images collected from diverse real-world text scenes, including street scenes, shop signs, billboards, posters, decorative text, art lettering, and magazine covers. The dataset includes Chinese text and a small amount of English text. Each image is annotated with text localization information and transcription labels, including polygon and quadrilateral bounding boxes for vertical and non-vertical text regions. This dataset is suitable for tasks such as scene text OCR tasks and multi-oriented text recognition.

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.