Why didn't you freeze any layers of VGG before training with your data ?. Is there any advantage ? How did you decide?

Can you help me regarding installation of the data_helper package. I'm not getting to import that package.

Thanks a lot for sharing this notebook! Can you share some info about the environment you're running it on? I'm having quite different and worst results that what is printed in the notebook when running it :(

The validation data is sampled in train_model and in the notebook train_model is called several times, therefore validating on data that was previously used for training:

train_losses, val_losses = [], []
epochs_arr = [20, 5, 5]
learn_rates = [0.001, 0.0001, 0.00001]
for learn_rate, epochs in zip(learn_rates, epochs_arr):
    tmp_train_losses, tmp_val_losses, fbeta_score = classifier.train_model(x_train, y_train, learn_rate, epochs, 
                                                                           batch_size, validation_split_size=validation_split_size, 
                                                                           train_callbacks=[checkpoint])
    train_losses += tmp_train_losses
    val_losses += tmp_val_losses

(

I don't think this is intentional so I'm raising the issue :)

When I run the amazon_forest_notebook.py, there is a syntax error.
Traceback (most recent call last):
File "amazon_forest_notebook.py", line 43, in
from keras_helper import AmazonKerasClassifier
File "../src/keras_helper.py", line 35
self.classifier.add(BatchNormalization(input_shape=(*img_size, img_channels)))
^
SyntaxError: invalid syntax.
Why do I have this error? Thank you!

Hi, trying to run amazon_forest_notebook.py I get the following error:

Traceback (most recent call last):
  File "amazon_forest_notebook.py", line 45, in <module>
    import data_helper
  File "/home/bio/anaconda3/lib/python3.6/site-packages/data_helper/__init__.py", line 3, in <module>
    import check
ModuleNotFoundError: No module named 'check'

How can I solved,
Thanks in advance,
Luis Alfonso

Use cross-entropy on all weather labels of the training data (add the None label to those who don’t have any).
The labels concerned are: clear, partly cloudy, cloudy, haze