In this recipe, you are going to learn how to use torchvision's pretrained (on Imagenet) deep learning models for a few famous models. ImageNet is an image database organized as per the WordNet hierarchy. Hundreds/thousands of images belong to each node in the hierarchy.
The following plot shows the top-1 accuracy achieved by a few popular deep neural nets participated in the ImageNet challenge, starting from AlexNet (Krizhevsky et al., 2012) on the far left, to the best performing Inception-v4 (Szegedy et al., 2016) on the far right:
The top-1 accuracy is defined as the average number of times the correct label for an image was the highest probability class predicted by the CNN for that image. At the other end of the scale, the top-1 error shows the error that occurs when the model-predicted...