如何在 tensorflow 数据集上使用 mobilenet_v2.preprocess_input

Question

I'm again struggling with the usage of tensorflow datasets. I'm again loading my images via

data = keras.preprocessing.image_dataset_from_directory(
  './data', 
  labels='inferred', 
  label_mode='binary', 
  validation_split=0.2, 
  subset="training", 
  image_size=(img_height, img_width), 
  batch_size=sz_batch, 
  crop_to_aspect_ratio=True
)

I want to use this dataset in the pre-trained MobileNetV2

model = keras.applications.mobilenet_v2.MobileNetV2(input_shape=(img_height, img_width, 3), weights='imagenet')

The documentation says, that the input data must be scaled to be between -1 and 1 ( https://www.tensorflow.org/api_docs/python/tf/keras/applications/mobilenet_v2/MobileNetV2 ). To do so, the preprocess_input function is provided. When I use this function on my dataset

scaled_data = tf.keras.applications.mobilenet_v2.preprocess_input(data)

I get the error: TypeError: unsupported operand type(s) for /=: 'BatchDataset' and 'float'

So how can I use this function properly with the tensorflow dataset?

Answer 1

Maybe try using tf.data.Dataset.map :

import tensorflow as tf
import pathlib

dataset_url = "https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz"
data_dir = tf.keras.utils.get_file('flower_photos', origin=dataset_url, untar=True)
data_dir = pathlib.Path(data_dir)

batch_size = 32

train_ds = tf.keras.utils.image_dataset_from_directory(
  data_dir,
  validation_split=0.2,
  subset="training",
  seed=123,
  image_size=(180, 180),
  batch_size=batch_size)

def preprocess(images, labels):
  return tf.keras.applications.mobilenet_v2.preprocess_input(images), labels

train_ds = train_ds.map(preprocess)

images, _ = next(iter(train_ds.take(1)))
image = images[0]
plt.imshow(image.numpy())

Before preprocessing the images:

After preprocessing the images with tf.keras.applications.mobilenet_v2.preprocess_input only:

After preprocessing the images with tf.keras.layers.Rescaling(1./255) only: