tesseract识别精度差

[mengbin890815]

1. tesseract 语言包是经过训练的吗，我自己的tesseract 识别很差，感觉demo的‘Login’ 都识别不到，跑的结果很差。自己平常工作中设计的ocr的，用tesseract精度也比较差。
   如果需要训练的话，可否有训练好的traineddata语言包提供下呢。
2. 根据配置的说明，去掉strategy: 'ocr' 这个参数的话，又提示terminate due to timeout。 NoSmoke 1.0 现在无法兼容了吗？
3. 截图灰度阈值在哪里设置？截图二值化后，我的好多截图，直接把有用的信息处理掉了。

[macaca-bot]

This is the translated issue:

---

Tesseract recognition accuracy difference

Is the Tesseract language package trained? My own tesseract is very bad, I feel that Demo's 'login' recognizes that the result of running is poor. The OCR designed in our usual work is also relatively poor with Tesseract.