Chinese internet giant Baidu has quietly released an optical character recognition (OCR) API that developers can use to convert image-based text into machine-encoded text.
Demand for data has led to an increase in the popularity of OCR APIs as companies seek to turn unstructured information, like images with text, into structured information that they can analyze.
Baidu's OCR API offers developers 5,000 free API requests per month for images up to 300 kilobytes in size. The current languages supported are Chinese, English, Japanese and Korean. Documentation for the OCR API is available in Chinese only. To use the API, a Baidu developer account, which is free, is required.
Already, there are two open source projects that help developers take advantage of Baidu's free offering. One uses Node.js and the other is implemented as a Visual Studio C# project. The latter has documentation in English, which could help English-speaking developers who are not able to access Baidu's Chinese documentation.
Baidu's OCR API is in beta. Given this and its usage limits, it might not yet be suitable for production use. Developers with commercial OCR needs can turn to a growing number of cloud OCR API solutions from companies like ABBY and Accusoft.