The dataset consists of 500,000 images for multi-country natural scenes and document OCR, including 20 languages such as Traditional Chinese, Japanese, Korean, Indonesian, Malay, Thai, Vietnamese, Polish, etc. The diversity includes various natural scenarios and multiple shooting angles. This set of data can be used for multi-language OCR tasks.
For more details, please refer to the link: https://www.nexdata.ai/datasets/speechrecog/1759?source=Github
500,000 images. For each language, there are 25,000 images in total, including 12,500 natural scene images and 12,500 document images
traditional Chinese, Japanese, Korean, Indonesian, Malay, Thai, Vietnamese, French, German, Italian, Portuguese, Russian, Spanish, Arabic, Turkish, Polish, Dutch, Greek, Czech, Filipino (Tagalog)
Natural scene: including slogan, receipt, poster, warning sign, road sign, food packaging, billboard, station sign and signboard, etc. Document: electronic documents, meeting minutes, reports, manuals, user manuals, books, newspapers, teaching materials, etc.
including a variety of natural scenes, multiple shooting angles
cellphone, scanner
looking up angle, looking down angle, eye-level angle
according to the collection requirements, the collection accuracy is not less than 97%
Commercial License