In September 2012, a team supervised by Dr. Geoffrey Hinton from the University of Toronto, considered the godfather of deep learning, competed to build AlexNet. AlexNet was training against a behemoth image test set called ImageNet. ImageNet consisted of more than 14 million images in over 20,000 different classes. AlexNet handily beat its competition, a non-deep learning solution, by more than 10 points that year and achieved what many thought impossible – that is, the recognition of objects in images done as well or perhaps even better than humans. Since that time, the component that made this possible – CNN – has in some cases surpassed human cognition levels in image recognition.
The component that made this possible, CNN, works by dissecting an image into features – features that it learns to detect by learning to detect those...
