SwiftOCR
SwiftOCR copied to clipboard
How to improve the recognition rate
How to improve the recognition rate I have some captcha images here
![captcha](https://user-images.githubusercontent.com/19711721/36100958-ebad5b1a-1042-11e8-98db-bb09ca553320.png)
I try to use SwiftOCR to identify these codes
This is the result of identification
![result](https://user-images.githubusercontent.com/19711721/36101076-468a2f18-1043-11e8-94c1-9eccd373d70d.png)
Here is a demo program OCR.zip
How should I improve the recognition rate? Or how should I remove these Unrelated lines
Improving accuracy for recognizing captchas won't be possible. SwiftOCR probably work properly with distorted text. What you can do is you can remove the unrelated lines and train SwiftOCR to also recognize lowercase character. If that doesn't help, you should search for another solution.
I have tried recognize lowercase character It does not help For removing unrelated lines Do you have the relevant processing experience to share?
You can try to apply an morphological closing filter to remove the lines but keep the characters.