CCPD

CCPD copied to clipboard

test error

1

Why does the test fail to display Chinese character but have english and number license plates?

dingning94

testing the demo on independent data (0% accuracy), any explanations?

1

Hey, So first of all i'm not interested in the recognition but just detection/localizatoin of the plate and i tried images of an independent dataset and demo.py failed badly as...

Moadab-AI

How to split the train and val set?

1

In order to compare the performance in the paper, can you elaborate on the image list of the 100k training set?

ljd16

不能做一张图片中有多车牌的检测和识别 对于没有车牌的图片也会输出检测结果 检测部分不支持旋转检测 识别部分不支持不定长字符的识别（新能源车8位）；使用了7个不同的分类器，模型参数多。

lanybass

RPnet classifier

1

Thank you for your excellent work.and there i have one question.When it comes to character recognition, you use seven different classifiers, and the next five categories should be the same....

chenjun2hao

皖开头的车牌占了训练集和测试集总数的90%以上，所以导致汉字识别能力较差。谁能共享下其他的训练集啊，谢谢。

rochesterlmg

Detection precision problem!

1

Excuse me, it seems that the eval.py you provide only measures the recognition precision. Or maybe I just miss the code when reading your code? Can you provide a code...

lcy0604

First at all，thanks for your kindly open CCPD source and datasets，it is a great work. I had downloaded the dataset from the provide link, but, I found that the provide...

sirius-ai

loss += 0.8 * nn.L1Loss().cuda()(fps_pred[:][:2], y[:][:2]) loss += 0.2 * nn.L1Loss().cuda()(fps_pred[:][2:], y[:][2:]) With reference to above lines? I understand you are taking .8 times loss of first two examples ??...

pshroff04

trainset: 100k images in ccpd_base folder testset: 20k images in ccpd_base folder from the 10th epoch to 100th epoch, the test accurate is Lingering around 95% Had I known what...

USTC-SQQ