search

sirius-ai / LPRNet_Pytorch

Pytorch Implementation For LPRNet, A High Performance And Lightweight License Plate Recognition Framework.
Apache License 2.0
936 stars 233 forks source link

省份识别精度不高 #57

Open duorouputao opened 3 years ago

duorouputao commented 3 years ago

有人遇到过车牌中文识别精度不高，但是英文和数字精度都还行的情况吗

littlePrince126 commented 3 years ago

I meet the same problem.

duorouputao commented 3 years ago

Well,U can change the dataset, and use another loss function.It will improve the performance of the network.

littlePrince126 commented 3 years ago

Hi! could you tell me which loss you used in the work? Thanks a lot!

At 2021-07-30 09:15:56, "duorouputao" @.***> wrote:

Well,U can change the dataset, and use another loss function.It will improve the performance of the network.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

lileicv commented 3 years ago

我怀疑这个模型是CCPD数据集训练的，那个数据集中大多数车辆都是安徽的，所以“皖”特别多，导致模型容易把省份错分成“皖”。

duorouputao commented 3 years ago

但是其他的也不行，最容易出现错误的就是车牌的首字符中文

lhh1005 commented 1 year ago

有人遇到过车牌中文识别精度不高，但是英文和数字精度都还行的情况吗

请问一下这个有好的解决办法嘛，我训练的loss值很低了，但是测试的时候识别效果不好，请问一下这种该咋调呀