Convolutional Neural Network

Until around 2015, image tasks such as face recognition were typically done by means of laborious hand coded programs that picked up facial features such as eyebrows and noses. Many OCR or face recognition applications were not using machine learning at all. The rapid acceleration in computing power, and the wide availability of large datasets, GPUs, and deep learning software, meant that around the mid 2010s, convolutional neural networks were able to deliver much better accuracy than the traditional methods and suddenly became the standard for nearly all computer vision related tasks in academia and industry. Now the average smartphone user probably has one or two apps running convolutional neural networks in their pocket, a concept that would have been unthinkable in 2010.