CCPD
CCPD copied to clipboard
detection precision problem
I use the code 'wR2.py' and the dataset you provided to train a detection model with 25 epochs.
Because I can not find the code to evaluate the precision of detection model,so I write it by myself.
But I find the license plate detection results are very poor.Why?
@detectRecog
wR2.out as follow:
I am also getting very poor results. I found a couple bugs that I fixed but the results are still poor (Roughly 82.8% on Val set). Did you manage to resolve?
I am also getting very poor results.23% in ccpd_db