Dog Breed Classification

Introduction

This project was my first project in Udacity's AI programming nano degree. We used pre-trained models such as ResNet, VGG, Alexnet, and our own designed architecture to check if an image is a picture of a dog and then recognize its breed.

Usage

The input images should be:

• In jpeg format with extension jpg.
• In approximately square in shape (their height and width are approximately the same numbers of pixels).
• If you want to check the accuracy, name the files like this: Animal_Name_01.jpg or Object_Name_01.jpg

Then Put all images you want to classify in the uploaded_images folder, open a terminal window in the project directory and type the following, then hit enter:

sh run_models_batch_uploaded.sh

To do this, you will be calling the following shell script that will output the following results files:

• resnet_uploaded-images.txt - that contains the results using CNN model architecture ResNet
• alexnet_uploaded-images.txt - that contains the results using CNN model architecture AlexNet
• vgg_uploaded-images.txt - that contains the results using CNN model architecture VGG

Results

In this project, we had two main objectives:

• Identifying which pet images are of dogs and which pet images aren't of dogs
• Classifying the breeds of dogs, for the images that are of dogs

In the table below, you will find our results for each of the model architectures.

• For objective 1, notice that both VGG and AlexNet correctly identify images of "dogs" and "not-a-dog" 100% of the time.
• For objective 2, VGG provides the best solution because it classifies the correct breed of dog over 90% of the time.

Conclusion

Given our results, the "best" model architecture is VGG. It outperformed both of the other architectures when considering both objectives 1 and 2. You will notice that ResNet did classify dog breeds better than AlexNet, but only VGG and AlexNet were able to classify "dogs" and "not-a-dog" at 100% accuracy. The model VGG was the one that was able to classify "dogs" and "not-a-dog" with 100% accuracy and had the best performance regarding breed classification with 93.3% accuracy.