CODEWITHSUNDEEP
tensorflowinputneural-networkmodelconv-neural-network
S.M
S.M

Reputation: 131

Tensorflow Model Data cardinality is ambiguous

I have a Tensorflow with the following layers model, I believe that the problem lies with the input shape of the first layer. all the images that i am using are colored and has 492x702 resolution.

# Initialize empty lists
images = np.array([])
labels = np.array([])

# read and convert each image to an array
# then append it to the images list and append the result label to the label list
for index, folder in enumerate(os.listdir(path)):
    print("Folder Index = " + str(index))
    if folder == ".DS_Store":
        continue
    for i, image in enumerate(os.listdir(path + '/' + folder)):
        print("Image Index = " + str(i))
        image_path = path + '/' + folder + '/' + image
        img = plt.imread(image_path)
        image_array = tf.keras.preprocessing.image.img_to_array(img)
        images = np.append(images, image_array)
        labels = np.append(labels, folder)

training_images = images
training_images = training_images / 255.0


# model structure
model = tf.keras.models.Sequential([
    tf.keras.layers.Conv2D(32, (3, 3), activation=tf.nn.relu, input_shape=(702, 492, 1)),
    tf.keras.layers.MaxPooling2D(2, 2),
    tf.keras.layers.Conv2D(64, (3, 3), activation=tf.nn.relu),
    tf.keras.layers.MaxPooling2D(2, 2),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(128, activation=tf.nn.relu),
    tf.keras.layers.Dense(100, activation=tf.nn.softmax)
])

# compile the mode
model.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(), metrics=['accuracy'])
model.summary()

# train the model
model.fit(training_images, labels, epochs=3)

Error:

ValueError: Data cardinality is ambiguous:

x sizes: 725306400

y sizes: 700

Make sure all arrays contain the same number of samples.

Am I doing something wrong?

Upvotes: 0

Views: 247

Answers (1)

mz2300
mz2300

Reputation: 489

Pay attention on images.shape. After applying np.append() you get a flat array instead of 3d array. So as a result you pass a flat array to NN input which expects an object having (702, 492, 1) shape

Upvotes: 1

Related Questions