Which of the Following Statements Is False

Multiple Choice

Which of the following statements is false?

A) The number of filters depends on the image dimensions-higher-resolution images have more features, so they require more filters.
B) If you study the code the Keras team used to produce their pretrained convnets, you'll find that they used 64, 128 or even 256 filters in their first convolutional layers. Based on their convnets and the fact that the MNIST images are small, we used 64 filters in our first convolutional layer.
C) The set of filters produced by a convolution layer is called a feature map. Subsequent convolution layers combine features from previous feature maps to recognize larger features and so on. If we were doing facial recognition, early layers might recognize lines, edges and curves, and subsequent layers might begin combining those into larger features like eyes, eyebrows, noses, ears and mouths.
D) Once the network learns a feature, because of convolution, it no longer needs to recognize that feature elsewhere in the image.