Getting the Error when running a image classification

TensorFlow General Discussion
, , ,
4

@vivek_yadav

The error you’re encountering, “AttributeError: ‘str’ object has no attribute ‘base_dtype’,” is likely due to a mismatch in data types when passing the labels to the fit function.

In your case, the issue might be related to the labels (y_train_one_hot and y_test_one_hot) being of type string.

Here are a few suggestions we have for you:

  1. Confirm that the data types of your labels (y_train_one_hot and y_test_one_hot) are appropriate. They should be of numeric types, preferably integers. Use astype to convert the label arrays to integers before passing them to the fit function.
  2. Since you are using one-hot encoding for your labels, make sure the label encoding step is done correctly.

I have tried to modify version of your code to address these points:

# Encode labels from text to integers.
from sklearn import preprocessing
le = preprocessing.LabelEncoder()
le.fit(test_labels)
test_labels_encoded = le.transform(test_labels)
le.fit(train_labels)
train_labels_encoded = le.transform(train_labels)

# Ensure one-hot encoding of labels
from keras.utils import to_categorical
y_train_one_hot = to_categorical(train_labels_encoded)
y_test_one_hot = to_categorical(test_labels_encoded)

# Convert labels to integers
y_train_one_hot = y_train_one_hot.astype(int)
y_test_one_hot = y_test_one_hot.astype(int)

# Train the CNN model
history = cnn_model.fit(x_train, y_train_one_hot, epochs=2, validation_data=(x_test, y_test_one_hot))

I have tried to label correctly as integers and then one-hot encoded. Also, it ensures that the label arrays have the correct data type before passing them to the fit function.

Try running the modified code, and it should resolve the AttributeError you’re facing. If you encounter any further issues, feel free to ask!