Closed ChillyGuy closed 7 years ago
That's because it didn't strongly match the first "9" character. It was more confident that it wasn't sure what it was:
Improving the quality of the OCR training set may help
PostProcess Line 0 Letter: 0 ۲ -- score: 137.066 -- occurrences: 6 PostProcess Line 0 Letter: 0 ۳ -- score: 9.24496 -- occurrences: 1 PostProcess Line 0 Letter: 0 ~ -- score: 5.75504 -- occurrences: 1 PostProcess Line 0 Letter: 1 ۲ -- score: 134.638 -- occurrences: 6 PostProcess Line 0 Letter: 2 و -- score: 73.7063 -- occurrences: 6 PostProcess Line 0 Letter: 2 ~ -- score: 33.014 -- occurrences: 5 PostProcess Line 0 Letter: 2 ۶ -- score: 1.49072 -- occurrences: 1 PostProcess Line 0 Letter: 3 ۵ -- score: 113.704 -- occurrences: 6 PostProcess Line 0 Letter: 4 ۳ -- score: 75.3051 -- occurrences: 6 PostProcess Line 0 Letter: 4 ~ -- score: 41.1455 -- occurrences: 10 PostProcess Line 0 Letter: 4 ۲ -- score: 28.5798 -- occurrences: 3 PostProcess Line 0 Letter: 4 ۴ -- score: 4.96964 -- occurrences: 1 PostProcess Line 0 Letter: 5 ~ -- score: 18.8806 -- occurrences: 2 PostProcess Line 0 Letter: 5 ۹ -- score: 11.1194 -- occurrences: 2 PostProcess Line 0 Letter: 6 ۹ -- score: 70.4581 -- occurrences: 6 PostProcess Line 0 Letter: 6 ~ -- score: 22.4862 -- occurrences: 4 PostProcess Line 0 Letter: 7 ۹ -- score: 62.8051 -- occurrences: 4 PostProcess Line 0 Letter: 7 ~ -- score: 4.51016 -- occurrences: 2
Hello Mat, Thanks for responding i'll try to improve ocr library.
Hello dear \@ChillyGuy , I m working on persian alpr , did u got any result , i need some help
Hey, Yep, it takes a bit time to get the hang of it but once you do you can get it to work.
hello dear @ChillyGuy
I'm working on a project about recognizing moroccan license plates which look like this image :
i did not have datasets , i decide to use "us" and "eu" alpr
i tried to modify "us.patterns" i added the line " mr #####@# "
When I run the command line " alpr -p mr lprmr.jpg" it give me this result :
plate0: 8 results
Hello! Im working on training alpr for Iranian plates, theres a problem which is bugging me ever since i started my custom ocr traineddata against plates.
for some reason some characters are not detectable by the program and I'm clueless how to fix it. as an example in screen shot below plate on the right has 8 characters but ALPR detects only 7 funny thing is it has detected it correctly but for some reason demoted the result to the second place ... (correct result has confidence : 79.8315)
debug.txt
Thanks in advance