This video looks at Convolutional Neural Networks 'CNNs'. These powerful networks are ideally suited to computer vision. They are able to look at local changes in data to pull out pertinent features such as edges. Further layers in the network can combine these edges into lines, and these lines into corners, shapes, patterns and textures. These more abstract features can then be fed into a fully connected layer for classification. The classification results from abstractions (patterns, shapes, lines textures etc.) are far superior to the classifications of raw pixel data alone.
These CNNs are at the heart of many modern applications of AI - medical diagnostics, biometrics and autonomous vehicles.