What does Conv2D(32, (3, 3) in TensorFlow mean?

Question

On the CNN image classification example from a tensor's flow tutorial page (https://www.tensorflow.org/tutorials/images/cnn),

There is a code that goes like

model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))

I understand that the image is 32 by 32 with a channel of 3 for RGB but what does the Conv2D(32, (3, 3) represent? Specifically the (3,3).

Answer 1

The (3,3) specifies the shape of the convolutional kernel. Check out the docs for more information.