Applications of Convolutional Neural Networks (CNNs)

More examples of deep networks

We have demonstrated the $\text{CNN}$ on the classic $\text{MNIST}$ dataset, but it’s interesting to note that these networks gained wide recognition in another dataset called $\text{ImageNet}$. $\text{ImageNet}$ is a collection of over $\text{one}$ million pictures from the Internet that have been labeled for over $\text{1,000}$ classes. This dataset has been used for a competition where some traditional machine learning methods such as SVMs have gained some success. However, the traditional methods stagnated, with accuracies remaining around $\text{80}$ percent. Alex Krizhevsky and colleagues applied $\text{CNNs}$ to the problem and won a competition in 2012, starting a trend of improving the performance of models in object recognition even further.