|
|
|
|
|
© Copyright 2004-2010
LPR Editor
|
|
Software for OEMs
LPReditor provides to its customers the possibility to adapt its engine to thier needs.
That technology ensures that you obtain better results.
That is the hearth of our engine.
Options:
You can do a basic, a medium or a complete or a partial retraining.
Requirements:
For the the basic or the medium retraining you need 1000 license plates pictures.
For a complete retraining you need 2500 license plates pictures.
The image filenames must contain the exact registration
(after a "_" (underscore) and at the end of the filename) like "005_WGENI31.jpg"
or "0000000044_2979TW34.bmp".
The basic retraining:
Easy to use
is fully automatic
is an application (LPRBasicRetrainer) that does not create new neural nets for your application but
just select the best neural nets for your application (among a wide collection of neural nets are made available by LPReditor)
A basic retraining will result in a better reading rate (at least 5 %)
The medium retraining:
is fully automatic
is an application (LPRMediumRetrainer) that creates new neural nets for your application
uses a database of more than 100 000 cut and classified characters
A medium retraining will result in a better reading rate (at least 8 %).
The complete retraining:
is an application (LPRFullRetrainer) that cuts and classifies each character
and stores it in a dedicated database.
This task is not fully automatic : You need to check carrefully that each character
is correctly classified.
After that you proceed with a medium retraining based on the new database
(for that you can reuse the same images files) to producenew neural nets.
A complete retraining will result in a better reading rate (at least 10 %).
|
|
|
