the label written in tfrecord is not same with the label read

MaybeShewill-CV / CRNN_Tensorflow

Convolutional Recurrent Neural Networks(CRNN) for Scene Text Recognition

MIT License

1.03k stars 388 forks source link

Closed 13438960761 closed 5 years ago

13438960761 commented 5 years ago

我写入tfrecord格式的label是索引值24,26,3,25,12 ；但是从tfrecord解析出来的label索引是2.25.8.14；不一样；图片路径是正确的；不知道是否与解析的稀疏tensor有关系？

MaybeShewill-CV commented 5 years ago

@13438960761 不知道你是怎么解析的我生成label和读取label是对应的没有问题。因为没有专门的测试单元测试这个所以你可以用小batch数据训练看模型是否能收敛：）