zonasw / dbnet Goto Github PK

谢谢！

Hi, may I ask which label tool you use? cause I want to label my own image~

I know it dose not impact the model just working, i just want konw why did you set this

in batch norm layer setting layer.trainable = False means freeze the layer, i.e. its internal state will not change during training: its trainable weights will not be updated during fit() or train_on_batch(), and its state updates will not be run，why batch norm freeze in train code?

您好，我测试了你开放的预训练模型，效果很好，我想自己训练下，wenmuzhou的检测提供了十几个数据集，您全用到了么？其中有很多弯曲文本的和阿拉伯文的数据集，想知道您训练用到了那几个数据集，我想复现训练下。

还有训练时，默认cpu训练么？如何修改为指定gpu啊，2.0的接口不太熟悉

The trained model has a better result in my test case, i want to know the details about trainging process, thanks!

Hi,
There are lots of dataset, which one was used for train?

 icdar2015
 MLT2019
 COCO-Text_v2
 ReCTS
 SROIE
 ArT
 LSVT
 Synth800k
 icdar2017rctw
 MTWI 2018
 百度中文场景文字识别
 mjsynth
 Synthetic Chinese String Dataset(360万中文数据集)
 英文识别数据大礼包

如题，您好，如何从多边形得到矩形？

Can this project achieve the same accuracy and speed as official implement?

你好，我在train.py中看到model.fit( x=train_generator, steps_per_epoch=cfg.STEPS_PER_EPOCH, initial_epoch=cfg.INITIAL_EPOCH, epochs=cfg.EPOCHS, verbose=1, callbacks=callbacks, validation_data=val_generator, validation_steps=cfg.VALIDATION_STEPS )
其中fit函数并没指定batch_size，那么STEPS_PER_EPOCH依旧是数据集的大小，并没有除以batch_size吗；但是在generator中却有指定了batch_size，也就是说数据集会除以batch_size。那么在训练的过程是否会出现在同一个epoch中，队某个样本进行重复训练？

训练时会时不时地报以下错误：
2020-12-06 16:33:17.421937: W tensorflow/core/kernels/data/generator_dataset_op.cc:103] Error occurred when finalizing GeneratorDataset iterator: Cancelled: Operation was cancelled
以及：
libpng warning: iCCP: known incorrect sRGB profile
libpng warning: iCCP: cHRM chunk does not match sRGB
加载的图片格式为JPG，但也有部分包含PNG的图片。
请问以上问题应该如何解决呢？

尝试将DBNet改成支持多卡，但是实践发现，只有使用tf.dataset读取数据时才能支持多卡，请问有没有这样的尝试呢？