How can i train tesseract from images directly ?

[Harshit Gupta]

I am having images from number plate of vehicles which isn't a standard font. I have cropped images of characters in number plate. I tried the following - 

1. Created a grid of all images so tesseract can read them for training and generated tif file for it.
2. Then i generated box files for the corresponding tif file and corrected the box file for wrongly detected characters.
3. Then i tried generating - unicharset, font_properties, shapeclustering, mftraining, cntraining.
4. At shapeclustering i am getting something like - 
   1. Bad properties for index 3, char ‘: 0,255 0,255 0,0 0,0 0,0
   2. Bad properties for index 4, char J: 0,255 0,255 0,0 0,0 0,0
   3. Bad properties for index 5, char l: 0,255 0,255 0,0 0,0 0,0
   4. Bad properties for index 6, char I: 0,255 0,255 0,0 0,0 0,0
   5. Bad properties for index 7, char .: 0,255 0,255 0,0 0,0 0,0

[ahmed.ba...@gmail.com]

I guess you use linux and you make trainng manualy , so

If you are within linux you can use LIOS is very efficase 

https://sourceforge.net/projects/lios/

When under windows , use jtestboxeditor and serak 

https://www.youtube.com/watch?v=47rgBL9NZkM