This project is part of Udacity's AI Programming with Python Nanodegree.

The goals of this project is to improve Python skills and to determine how well the "best" classification algorithm works on correctly identifying a dog's breed.

• Anaconda 5.3 with Python 3

Number of Images : 40
Number of Dog Images : 30
Number of Non-Dog Images : 10
% Correct Dogs : 1.00
% Correct "Not-a" Dog : 1.00
% Correct Breed : 0.93
% Match : 0.88

Misclassified Breed's of Dog:
Pet image: great pyrenees, Classifier label: kuvasz
Pet image: beagle, Classifier label: walker hound, walker foxhound

Total Elapsed Runtime: 0:0:51

Number of Images : 40
Number of Dog Images : 30
Number of Non-Dog Images : 10
% Correct Dogs : 1.00
% Correct "Not-a" Dog : 0.90
% Correct Breed : 0.90
% Match : 0.82

Misclassified Dogs:
Pet image: cat, Classifier label: norwegian elkhound, elkhound

Misclassified Breed's of Dog:
Pet image: great pyrenees, Classifier label: kuvasz
Pet image: beagle, Classifier label: walker hound, walker foxhound
Pet image: golden retriever, Classifier label: leonberg

Total Elapsed Runtime: 0:0:7

Number of Images : 40
Number of Dog Images : 30
Number of Non-Dog Images : 10
% Correct Dogs : 1.00
% Correct "Not-a" Dog : 1.00
% Correct Breed : 0.80
% Match : 0.75

Misclassified Breed's of Dog:
Pet image: great pyrenees, Classifier label: kuvasz
Pet image: beagle, Classifier label: walker hound, walker foxhound
Pet image: boston terrier, Classifier label: basenji
Pet image: beagle, Classifier label: english foxhound
Pet image: golden retriever, Classifier label: afghan hound, afghan
Pet image: golden retriever, Classifier label: tibetan mastiff

Total Elapsed Runtime: 0:0:3

The results are also presented in files vgg_uploaded-images.txt, resnet_uploaded-images.txt, alexnet_uploaded-images.txt.
The results of 4 images test are presendted in uploaded_images/my_classify_uploaded_images_results.txt.

According to the received results the best model architecture is VGG. VGG classified the provided dog images and uploaded images better since the results of classification were closer to the actual breeds and classification results of dog and non-dog images were done with 100% accuracy.