CODEWITHSUNDEEP
pythontensorflowkeras
jl303
jl303

Reputation: 1609

Randomly Transforming Everything in ImageDataGenerator at Once in tf.keras?

I'm using tensorflow.keras.utils.Sequence to model.fit_generator. I'm retrieving data and shuffling one batch at a time instead of loading everything into ram. In my __init__, I have self.datagen = ImageDataGenerator(width_shift_range=0.2, height_shift_range=0.2, zoom_range=0.2). Then in my __getitem__, I have:

self.datagen.fit(x_batch)
x_batch = next(self.datagen.flow(x_batch, batch_size=len(x_batch)))

Is this the best way to transform everything at once?

Upvotes: 1

Views: 106

Answers (1)

Hemerson Tacon
Hemerson Tacon

Reputation: 2522

You could just call fit_generator instead of fit and next. In this way, you wouldn't need to iterate over all your data. For more information about fit_generator take a look into keras help

Upvotes: 1

Related Questions