这东西挺好玩的，当年我还帮它修了个 memory leak 的 bug。

1楼 @frank4565 你厉害，中文识别的效果如何？

这东西据说不能识别手写？

Tesseract的中文识别率貌似很低 对英文印刷体识别率不错

不过好像是可以训练的

2楼 @tinyfool 不敢当不敢当，我修的只是这个wrapper。 它自带的那些model都只能识别最基本的，关键还是在模型训练上面。手写体肯定得自己训练了。

4楼 @frank4565 有点意思

Of course, you could use open source OCR engine, like Tesseract, Ocropus, and etc. If you like, it is very easy to design it by yourself using library about image processing and machine learning. OpenCV is a popular lib. I used to write a app about number plate recognition on my capstone. Here are some useful books with teaching codes.

Mastering OpenCV with Practical Computer Vision Projects

Instant OpenCV for iOS

6楼 @xuewenyuan 这个玩法我也想过，不着急，一个一个玩