sergiomsilva / alpr-unconstrained

License Plate Detection and Recognition in Unconstrained Scenarios

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Reproduction of experimental results

amaurylekens opened this issue · comments

I tried to reproduce the results for the OpenALPR datasets.For the European plates I obtained a score (percentage of correctly recognized plates) of 80.41% (93.52% in the paper) and for the Brazilian plates I obtained a score of 53.91% (91.23% in the paper) using only one class for 1 and i.

Considering O and 0 as a single class, I obtained a score of 97.94% for European plates and a score of 93.04% for Brazilian plates.

Does anyone get the right results? and should 0 and O generally be considered as one class ?

can you provide the script of evaluating the accuracy ?? and is it for the whole project or for the lp detection network....
because I'm trying to figure out one for lp detection

can you provide the script of evaluating the accuracy ?? and is it for the whole project or for the lp detection network....
because I'm trying to figure out one for lp detection

Hi Kenwann,

I am also working on upto License Plate detection and currently trying to calculate the model metrics only for License Plate detection.

Probably, If you are interested we can get in touch directly. Here is my mail id: hab9205@gmail.com

Thanks
Harsha