CODEWITHSUNDEEP
pythontensorflowconv-neural-networkclassificationimage-classification
black_powr
black_powr

Reputation: 11

Errors in Multi class image classification using CNN

I'm fairly new to CNN and python and I'm trying to classify images consisting of 3 classes. Whenever I try to train my model I get the error: TypeError: Cannot convert 0.6 to EagerTensor of dtype int64. I'm not fully sure what to make of this or how to fix it and would appreciate any help.

Here's the code :

train_gen=ImageDataGenerator(rescale=1./255)

validation_gen=ImageDataGenerator(rescale=1./255)

train_generator=train_gen.flow_from_dataframe(
    train_df,
    "/Users/danieladegborioye/Documents/school/FIFTH YEAR/ENGR 418 Applied Machine learning /Lego_dataset_1/training/",
    x_col='filename',
    y_col='category',
    target_size=(150,150),
    class_mode="categorical",
    batch_size=32
)

validation_generator=validation_gen.flow_from_dataframe(
    validate_df,
    "/Users/danieladegborioye/Documents/school/FIFTH YEAR/ENGR 418 Applied Machine learning /Lego_dataset_1/training/",
    x_col='filename',
    y_col='category',
    target_size=(150,150),
    class_mode="categorical",
    batch_size=32
)


epochs=10
batch_size=10

history=modell.fit(
    train_generator,
    epochs=epochs,
    validation_data=validation_generator,
    validation_steps=total_validate/batch_size,
    steps_per_epoch=total_train/batch_size,
    callbacks=callbacks
)

Upvotes: 1

Views: 144

Answers (1)

user11530462
user11530462

Reputation:

If batch_size is greater than the total_validate or total_train, then steps_per_epoch will be less than 1 which is causing this 'TypeError: Cannot convert 0.6 to EagerTensor of dtype int64'.

Upvotes: 0

Related Questions