What filter does TensorFlow use by default for image filtering?

Question

I'm new at machine learning and I was reading about CNN with Tensorflow but I have a doubt with this block of code:

conv1 = tf.layers.conv2d(
        inputs=input_layer,
        filters=32,
        kernel_size=[5, 5],
        padding="same",
        activation=tf.nn.relu)

According to several articles, there are a lot of filters to apply (Blur, Sharpening, etc). What kind of kernel is Tensorflow applying by default in this kind of code?

Because I'm not specifying any kernel type like:

I'm pretty confused about it, I'll really appreciate any kind of help.

Thanks in advance!

Answer 1

It's confusing, but TensorFlow has two conv2d methods: tf.nn.conv2d and tf.layers.conv2d. If you want to filter an image with a known kernel, call tf.nn.conv2d. If you want to create a layer in a convolutional neural network (CNN) that will determine its filters programmatically, call tf.layers.conv2d.