Hands-On ROS for Robotics Programming

Deep learning and neural networks

From now on, we will base our explanations on the practical example of the recognition of objects in images, which, in the case of the robot, will be supplied by the Raspberry Pi camera. In the following diagram, you can see a representation of a neural network that differentiates the three kinds of layers that there can be:

• The input layer is where we feed the dataset. Remember that such data has to be structured according to the selected features, that is, one neuron per feature. We will later discuss this particular and very common case of image datasets.
• The hidden layer(s)—one or more—are the intermediate steps in the deep learning pipeline that extract more features so that the algorithm is more capable ...