Dog Breed Classifier

Meeting Times: TBD

Room: BS 405, Florida Atlantic University

Meet the Team:

Alex Clark – Team Leader (


About the Project:

This project is to create a Neural Network that can identify the breed of a dog using only a photograph.  This project has advanced in stages as the work has been done to increase accuracy performance on the network on a 5-breed classification task.  We have tried three of the most prominent network architectures (AlexNet, VGG16, and ZFnet) as well as tweaked parameters to improve accuracy on the 5-breed task.  Accuracy levels have gone up from low 40% range (Spring 2017) to 95% (Spring 2018).

The next stages in this project involve: expanding to a wider range of dog breeds (the 50 most popular breeds) and attempting to achieve at least a 90% accuracy on that task, and (once this is achieved) making this network into an app that can be used by the public.  After this has been achieved, we will continue to improve the accuracy of the network by working with wrongly classified images.  We will also attempt to expand to an even wider range of dog breeds (perhaps over 100 breeds).