CODEWITHSUNDEEP
pythontensorflowtensorflow2.0
AdemxKartal
AdemxKartal

Reputation: 77

Error: Train inception v3 in tensorflow 2

I am new to tensorflow (version 2.3.0). I want to build an image classfier based on the 'oxford_flower102' with inception v3. I have prepared the dataset and now wants to train the inception v3 network, but I get an error which I do not understand. The error code is:

ValueError: Input 0 of layer conv2d is incompatible with the layer: : expected min_ndim=4, found ndim=3. Full shape received: [500, 667, 3]

The preprocess of the datasets seems to be fine, the error comes when I want to feed the data into the inception v3 network with the command

predictions = model(images, training=True)

Here you can find my whole code:

import warnings
import tensorflow as tf
import tensorflow_datasets as tfds
tfds.disable_progress_bar()
import logging
logger = tf.get_logger()
logger.setLevel(logging.ERROR)

import os
import tensorflow as tf
warnings.filterwarnings('ignore')
EPOCHS = 50
BATCH_SIZE = 8
NUM_CLASSES = 102
image_height = 299
image_width = 299
channels = 3
save_model_dir = os.getcwd()


def get_model():
    model = tf.keras.applications.InceptionV3(include_top=True,weights=None,classes=NUM_CLASSES)
    model.build(input_shape=(None, image_height, image_width, channels))
    return model

def get_loss_object():
    return tf.keras.losses.SparseCategoricalCrossentropy()


def get_optimizer():
    return tf.keras.optimizers.Adadelta()

def get_train_loss():
    return tf.keras.metrics.Mean(name='train_loss')

def get_train_accuracy():
    return tf.keras.metrics.SparseCategoricalAccuracy(name='train_accuracy')

def get_valid_loss():
    return tf.keras.metrics.Mean(name='valid_loss')
def get_valid_accuracy():
    return tf.keras.metrics.SparseCategoricalAccuracy(name='valid_accuracy')

if __name__ == '__main__':
    dataset, dataset_info = tfds.load('oxford_flowers102', with_info=True, as_supervised=True)
    dataset_info

    test_set, training_set, validation_set = dataset['test'], dataset['train'], dataset['validation']

    num_training_examples = 0
    num_validation_examples = 0
    num_test_examples = 0

    for example in training_set:
        num_training_examples += 1

    for example in validation_set:
        num_validation_examples += 1

    for example in test_set:
        num_test_examples += 1



    model =get_model()
    loss_object = get_loss_object()
    optimizer=get_optimizer()
    train_loss= get_train_loss()
    train_accuracy=get_train_accuracy()
    valid_loss=get_valid_loss()
    valid_accuracy=get_valid_accuracy()

    @tf.function
    def train_step(images, labels):
        with tf.GradientTape() as tape:
            predictions = model(images, training=True) #here I get the error

    # start training
    for epoch in range(EPOCHS):
        train_loss.reset_states()
        train_accuracy.reset_states()
        valid_loss.reset_states()
        valid_accuracy.reset_states()
        step = 0
        for images, labels in training_set:
            step += 1
            train_step(images,labels)

Upvotes: 1

Views: 337

Answers (2)

Guillem
Guillem

Reputation: 2647

Keras models expect batches of instances as input. You are feeding images one by one. If you want to train a model I recommend you batching the dataset. You can do it as follows:

dataset, dataset_info = tfds.load('oxford_flowers102', with_info=True, as_supervised=True)
    dataset_info

def resize_im(im, label):
  return tf.image.resize(im, [224, 224]), label

test_set, training_set, validation_set = dataset['test'], dataset['train'], dataset['validation']

training_set = training_set.map(resize_im)
test_set = test_set.map(resize_im)
validation_set = validation_set.map(resize_im)

training_set = training_set.batch(32)
test_set = test_set.batch(32)
validation_set = validation_set.batch(32)

# Then feed the images to the model

for images, labels in training_set:
  train_step(images, labels)

Reference: https://www.tensorflow.org/tutorials/load_data/images

Upvotes: 0

J.C.
J.C.

Reputation: 69

The shape of input is [none, width, height, channels], which are incomnpatible with your [500, 667, 3] apparently.

In the block of "start training", you are trying to use train_step for each image and label. Note that :

for images, labels in training_set:
    pass

indeed, you get image other than images:

for image, label in training_set:
    pass

You'd better use multiple images and labels for training, if you realy want use one image for each epoch, just reshape the image and the label.

For example, [500, 667, 3] to [1, 500, 667, 3] if you want use one image for a epoch.

Upvotes: 1

Related Questions