Convolution over volume – 1 x 1 filter

In this section, we will learn the significance of a 1 x 1 convolution and its use cases. A 1 x 1 convolution filter is a straight multiple of the image, as shown in the following diagram:

In the preceding diagram, the 1 x 1 convolution filter value of 1 is being used on a 5 x 5 image output from the previous section, but it can be any number in practice. Here, we can see that the use of a 1 x 1 filter conserves its height and width, whereas the depth increases to the filter channel's number. This is the fundamental advantage of a 1 x 1 filter. Each of the three (1 x 1 x 3) cells in the three-dimensional ...

Get Mastering Computer Vision with TensorFlow 2.x now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.