Deep learning is one element of diverse machine learning techniques that utilize artificial neural networks (ANN). The deep learning for vision systems can be supervised, semi-supervised, and unsupervised. Deep learning is a learning method carried out by machines by imitating how the basic systems of the human brain work. The basic system the human brain works on is called neural networks. That is why deep learning is called using artificial neural networks which in other words uses ‘artificial neural networks. Deep learning is a technology used in image recognition and computer vision.
In traditional machine learning, if we present a picture of a cat and then we ask the machine whether it is a cat or not, the machine’s ‘thinking’ process is based on the algorithm we created. For example, does an object have two eyes? Does the object have four legs? Does the object have long whiskers? Does the object have thick fur? If most or all of the answers are ‘yes’ then the machine will decide that it is a picture of a cat. Then what if the machine is presented with a picture like the one below? If we determine the learning algorithm as above, surely the answer from the machine is ‘no’.
Even though we as humans know that the picture is a picture of a cat. For example, we are asked to describe how we know if it is a cat from the specific characteristics of the object and then make the algorithm, of course, it will not work. Not all furry animals have tails, and ears dangling upwards are cats. Then how do we recognize it’s a cat? That’s when neural networks reach in.
Artificial Neural Networks commonly abbreviated as ANN is the most magical part of deep learning. This ANN simulates the work of our brain which is composed of neural networks called neurons. Just like the human brain system, in this artificial neural network, the machine receives information at points called nodes that are collected in one layer and then forwarded and processed to the next layer called hidden layers.