Baidu launches free OCR API (and nobody noticed?)
On June 1, 2015 Baidu -the biggest search engine in China- launched a new cloud service: Free OCR API (百度OCR文字识别). The catch? The documentation is in Chinese only. Thus, even so the launch was already three months ago, I could not locate a single English language blog post about it. This is a pity, it is really useful. So here we go.
The facts:
- To use the service, you need to sign up for a Baidu developer account (free, but in Chinese only.. you get the idea). The good news is that international (non-Chinese) mobile phone numbers work just fine for receiving the confirmation code.
-
The OCR API’s status is “experimental”/beta.
-
The use is free for up to 5000 requests/month. This is the second best offer in the market. The best one is, well…., our own OCR API with 15000+ free OCR conversions/month.
-
Supported OCR languages are (only) English, Korean, Japanese and Chinese.
-
Image size is limited to (only) 300kb, image format is JPG only.
-
The most interesting part: The OCR quality is very good, but not breathtaking. It plays in the same league as ABBYY and our own free ocr cloud service (which is powered by Microsoft OCR on the backend)
Getting started:
Jeremy Wei created a sample project that shows how to call Baidu OCR from Node.JS. For this review I created & published a Visual Studio sample project (C#) for testing the Baidu OCR with a C# GUI. Both projects are available on GitHub (MIT license).
The OCR test app in action:
Update: Several other blogs picked up on this news:
- Programmable Web: BaiDu offers Free OCR API
- CBN - China Business News: Baidu Inc Releases Free OCR API For Developers
- Mashup.se: API Måndag – WordPress, Google Spreadsheets API, Baidu OCR