I studied the differences between traditional image classification methods and deep learning methods. The test datasets include fashion-MNIST, CIFAR-10 and CIFAR-100.

The traditional method used are the bag of visual words, support vector machine and random forest classifier. The deep learning methods used are VGG13, VGG16 and VGG19

I found that in more complex image classification tasks, such as CIFAR100, which has 100 labeled categories, deep learning outperforms traditional machine vision methods by a large margin, with Bag of Visual Words performing the worst.