Image classification with Keras

Image Modeling with Keras

Ariel Rokem

Senior Data Scientist, University of Washington

Keras for image classification

from keras.models import Sequential

model = Sequential()

Keras for image classification

from keras.layers import Dense

train_data.shape

(50, 28, 28, 1)

Keras for image classification

model.add(Dense(10, activation='relu', 
          input_shape=(784,)))

model.add(Dense(10, activation='relu'))

model.add(Dense(3, activation='softmax'))

Keras for image classification

model.compile(optimizer='adam',         
              loss='categorical_crossentropy', 
              metrics=['accuracy'])

Keras for image classification

train_data = train_data.reshape((50, 784))

Keras for image classification

model.fit(train_data, train_labels, 
          validation_split=0.2,
          epochs=3)

model.fit(train_data, train_labels, 
          validation_split=0.2,
          epochs=3)

Train on 40 samples, validate on 10 samples
Epoch 1/3
32/40 [=========>......] - ETA: 0s - loss: 1.0117 - acc: 0.4688
40/40 [================] - 0s 4ms/step - loss: 1.0438 - acc: 0.4250 
                           - val_loss: 0.9668 - val_acc: 0.4000
Epoch 2/3
32/40 [=========>......] - ETA: 0s - loss: 0.9556 - acc: 0.5312
40/40 [================] - 0s 195us/step - loss: 0.9404 - acc: 0.5750 
                         - val_loss: 0.9068 - val_acc: 0.4000
Epoch 3/3
32/40 [=========>......] - ETA: 0s - loss: 0.9143 - acc: 0.5938
40/40 [================] - 0s 189us/step - loss: 0.8726 - acc: 0.6750 
                         - val_loss: 0.8452 - val_acc: 0.4000

Keras for image classification

test_data = test_data.reshape((10, 784))
model.evaluate(test_data, test_labels)

10/10 [==============================] - 0s 335us/step
[1.0191701650619507, 0.4000000059604645]

Let's practice!

Image Modeling with Keras