Identifying which pet images are of dogs (even if breed is misclassified) and which pet images aren't of dogs.

Classifying the breed of dog, for the images that are of dogs.

Determining which CNN model architecture (ResNet, AlexNet, or VGG), "best" achieve the objectives 1 and 2.

Considering the time resources required to best achieve objectives 1 and 2, and determine if an alternative solution would have given a "good enough" result, given the amount of time each of the algorithms take to run.