Tweaking your convolutions

Image Modeling with Keras

Ariel Rokem

Senior Data Scientist, University of Washington

Convolution

Convolution with zero padding

Zero padding in Keras

model.add(Conv2D(10, kernel_size=3, activation='relu', 
                 input_shape=(img_rows, img_cols, 1)),
                 padding='valid')

Zero padding in Keras

model.add(Conv2D(10, kernel_size=3, activation='relu', 
                 input_shape=(img_rows, img_cols, 1)),
                 padding='same')

Strides

Strides in Keras

model.add(Conv2D(10, kernel_size=3, activation='relu', 
                 input_shape=(img_rows, img_cols, 1)),
          strides=1)

Strides in Keras

model.add(Conv2D(10, kernel_size=3, activation='relu', 
                 input_shape=(img_rows, img_cols, 1)),
          strides=2)

Example

Calculating the size of the output

$$O = ((I - K + 2P) / S) + 1$$

where

• $I$ = size of the input
• $K$ = size of the kernel
• $P$ = size of the zero padding
• $S$ = strides

Calculating the size of the output

$$ 28 = ((28 - 3 + 2) / 1) + 1$$

$$ 10 = ((28 - 3 + 2) / 3) + 1$$

Dilated convolutions

Dilation in Keras

model.add(Conv2D(10, kernel_size=3, activation='relu', 
                 input_shape=(img_rows, img_cols, 1)),
          dilation_rate=2)

Let's practice!

Image Modeling with Keras