Skip to main content

Conv2D Layer

Summary​

  • filter
  • kernel size
  • padding
  • Strides

Content​

Conv2D Layer Example​

tf.keras.layers.Conv2D(
    filters=10,
    kernel_size=3,
    padding="valid",
    strides=(1, 1),
    activation="relu",
    input_shape=(224, 224, 3),
)

Filter​

The number of convolutional filters in that layer. In the example above, there are 10 filters. So, each filter will get a kernel weight of size 3 x 3 that's unique to each convolutional filter.

Kernel​

Determines the shape of the filter matrix (sliding window)

Padding​

Strides​

Number of steps a filter takes across an image at a time