I'm trying to make a basic MLP example in keras. My input data has the shape train_data.shape = (2000,75,75) and my testing data has the shape test_data.shape = (500,75,75). 2000 and 500 are the numbers of samples of training and test data (in other words, the shape of the data is (75,75), but there are 2000 and 500 pieces of training and testing data). The output should have two classes.
I'm unsure what value to use for the input_shape parameter on the first layer of the network. Using the code from the mnist example in the keras repository, I have (updated):
from six.moves import cPickle
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation
from keras.utils import np_utils
from keras.optimizers import RMSprop
# Globals
NUM_CLASSES = 2
NUM_EPOCHS = 10
BATCH_SIZE = 250
def loadData():
fData = open('data.pkl','rb')
fLabels = open('labels.pkl','rb')
data = cPickle.load(fData)
labels = cPickle.load(fLabels)
train_data = data[0:2000]
train_labels = labels[0:2000]
test_data = data[2000:]
test_labels = labels[2000:]
return (train_data, train_labels, test_data, test_labels)
# Load data and corresponding labels for model
train_data, train_labels, test_data, test_labels = loadData()
train_labels = np_utils.to_categorical(train_labels, NUM_CLASSES)
test_labels = np_utils.to_categorical(test_labels, NUM_CLASSES)
print(train_data.shape)
print(test_data.shape)
model = Sequential()
model.add(Dense(512, input_shape=(5625,)))
model.add(Activation('relu'))
model.add(Dropout(0.2))
model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dropout(0.2))
model.add(Dense(2))
model.add(Activation('softmax'))
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer=RMSprop(),
metrics=['accuracy'])
history = model.fit(train_data, train_labels, validation_data=(test_data, test_labels),
batch_size=BATCH_SIZE, nb_epoch=NUM_EPOCHS,
verbose=1)
score = model.evaluate(test_data, test_labels, verbose=0)
print('Test score:', score[0])
print('Test accuracy:', score[1])
where 5625 is 75 * 75 (emulating the MNIST example). The error I get is:
Error when checking model input: expected dense_input_1 to have 2 dimensions, but got array with shape (2000, 75, 75)
Any ideas?
From keras MLP example, https://github.com/fchollet/keras/blob/master/examples/mnist_mlp.py
# the data, shuffled and split between train and test sets
(X_train, y_train), (X_test, y_test) = mnist.load_data()
X_train = X_train.reshape(60000, 784)
X_test = X_test.reshape(10000, 784)
And the model input
model = Sequential()
model.add(Dense(512, input_shape=(784,)))
So you should reshape your train and test to (2000,75*75) and (500,75*75) with
train_data = train_data.reshape(2000, 75*75)
test_data = test_data.reshape(500, 75*75)
and then set the model input shape as you did
model.add(Dense(512, input_shape=(75*75,)))