Optical Character Recognition

Stable and reliable high-precision OCR service

Based on Tencent's deep learning technology and massive amounts of data, Tencent Cloud Optical Character Recognition (OCR) provides printed and handwritten character recognition services for a wide variety of cards, certificates and documents and supports custom templates, greatly improving data entry efficiency and reducing the cost of use.

Benefits

High Accuracy

Tencent Cloud Optical Character Recognition (OCR) can automatically locate and recognize text in images. It features robustness and an average accuracy rate of above 95% for printed text and 90% for handwritten text.

High Versatility

Developed independently by Tencent YouTu Lab, OCR covers all core algorithms for identity document analysis and recognition. It supports both landscape and portrait modes, and is applicable to scenarios with perspective distortion, irregular illumination, partial occlusion and more.

Ease of Use

OCR not only provides developers with a full range of APIs that can be called directly, but also SDKs that are highly compatible and easy to use.

Features

General Print Recognition

It supports the detection and recognition of text in images and features great versatility:
- It can recognize text in 19 languages, including Chinese, English, Japanese, Korean, Spanish, French, German, and Portuguese. It also supports text recognition in a supported language mixed with English.
- It can automatically detect the language, and return the text and bounding box coordinates.
- It supports automatic skew correction.

General Print Recognition (High-Precision)

It supports the detection and recognition of text in images and features higher accuracy:
- It can recognize Chinese text, English text, Chinese/English text, numbers, and special symbols with higher accuracy.
- It can recognize complex text with higher accuracy and recall, making it suitable for scenarios with a large amount of text, long numeric strings, small font, blurry or skewed text, etc.

Malaysian ID Card Recognition

It can quickly recognize key information on Malaysian ID cards:
- It can quickly recognize the text and extract structured fields, including ID number, name, gender, and address.
- It can recognize different types of ID cards, including MyKad, MyPR, MyTentera, MyKAS, POLIS, and IKAD.
- It can be used to crop passport photos.
- It can trigger alarms for spoofed, photocopied, and covered ID photos.

Hong Kong (China) ID Card Recognition

It can quickly recognize key information on Hong Kong (China) ID cards:
- It can quickly recognize the text and extract structured fields, including names in Chinese and English, telegraph codes, date of birth, gender, date of first issue, and ID number.
- It can be used to identify permanent residents for data archiving.

Passport Recognition

It can recognize key information on passports issued by Hong Kong/Macao/Taiwan (China), and other countries or regions:
- It can quickly recognize the text and extract structured fields, including passport ID, name, date of birth, gender, expiration date, issuing country/region, and nationality.
- It can be used to crop passport photos.
- It can trigger alarms for spoofed, photocopied, and covered ID photos.

General Print Recognition
General Print Recognition (High-Precision)
Malaysian ID Card Recognition
Hong Kong (China) ID Card Recognition
Passport Recognition

Scenarios

Identity Recognition in Finance

ID card recognition can be used in banking, insurance, securities, and other scenarios that require identity verification. It can reduce data entry workload, increase efficiency and improve the user experience.

Identity Recognition in the Internet Industry

In scenarios such as live streaming and e-hailing, OCR can verify the identities of hosts and passengers to reduce the risk of violations and crimes.

Identity Recognition
Passport Recognition
Inappropriate Text Recognition
Image-to-Text for Publications