aaronshan / 12306-captcha Goto Github PK

在训练模型的时候accuracy = 0.93， 但是测试的时候成功率特别低，差不多每个分类都是平均的概率，你们有没有遇到过这个情况啊。文字和图像部分的模型时一样的吗

请问作者您最终测试的准确率能达到多少？

tools 下面 words和remove_noisy都没找到

可不可以这样。比如：提示“档案袋”，如果一张图在“档案袋”多次出现，就给这张图加入‘档案袋’标签

老哥，你有区分测试集和训练集吗？还是直接全部直接用来训练？

hello,
我按照您的步骤训练后，船舰模型的时候不成功报一下错误，想请教下这个是怎么解决法，多谢
I0514 15:33:19.444087 3925 net.cpp:84] Creating Layer vgg-train-data
I0514 15:33:19.444136 3925 net.cpp:380] vgg-train-data -> data
I0514 15:33:19.444213 3925 net.cpp:380] vgg-train-data -> label
F0514 15:33:19.444285 3925 data_transformer.cpp:465] Check failed: datum_channels > 0 (0 vs. 0)
*** Check failure stack trace: ***
@ 0x7fcbcd0c3daa (unknown)
@ 0x7fcbcd0c3ce4 (unknown)
@ 0x7fcbcd0c36e6 (unknown)
@ 0x7fcbcd0c6687 (unknown)
@ 0x7fcbcd5aac4b caffe::DataTransformer<>::InferBlobShape()
@ 0x7fcbcd52bb25 caffe::DataLayer<>::DataLayerSetUp()
@ 0x7fcbcd585d64 caffe::BasePrefetchingDataLayer<>::LayerSetUp()
@ 0x7fcbcd4a3cf5 caffe::Net<>::Init()
@ 0x7fcbcd4a5c02 caffe::Net<>::Net()
@ 0x7fcbcd4af126 caffe::Solver<>::InitTrainNet()
@ 0x7fcbcd4b0183 caffe::Solver<>::Init()
@ 0x7fcbcd4b045f caffe::Solver<>::Solver()
@ 0x7fcbcd487241 caffe::Creator_SGDSolver<>()
@ 0x40d6a9 caffe::SolverRegistry<>::CreateSolver()
@ 0x407f5d train()
@ 0x405c5c main
@ 0x7fcbcc0cbf45 (unknown)
@ 0x4064cb (unknown)
@ (nil) (unknown)

谢谢O(∩_∩)O~