Hands-On ROS for Robotics Programming

The output layer

For the question of our example, that is, there is a cat in the picture? yes if the image shows a cat, or not if it doesn't. So we only need a neuron in the output layer, as shown in the diagram below. Then, if trained with many photos of cats, this network could classify an image to say whether it contains a cat (1) or not (0). An important point here is that the model should be able to identify the cat whatever position it occupies in the image, center, left, right, top, down, and so on:

Source: https://commons.wikimedia.org/wiki/File:NeuralNetwork.png

If we need to classify 10 kinds of objects (several types of pets, for ...