Dog Breed Classification.

Interactive Web Applications for Data Analysis.

Date. 2022

Team. Zona Kostic, Rob Dotson, and Gabor Korodi as supervisors.

Katerina Weikert and Connor Poetzinger as accomplices.

Tool. Figma, Photoshop, Illustrator

Dev.  Python, HTML/CSS/JS, Flask, Jinja, SQLAlchemy, CNN model, Pytorch, ResNet18

This project aims to create a web application that classifies dogs according to their breed. The user can access the built-in camera or download an image from his computer. The application then returns the first five breeds of dogs that match this search and their probability percentage.

For what?

It's aimed at city dwellers looking for specific breeds compatible with their current dog or simply out of curiosity. It could be a professional dog-sitter who wants to ensure his dogs will get along well or a simple owner looking for an ideal companion for a walk in the city. The app could also be helpful to prospects at dog shelters who want to be intelligently informed about the specific needs of a particular breed (or a mix of breeds) before acquiring an abandoned dog.

Technologies

The application was developed in Flask, Jinja, SQLAlchemy, and published on Heroku. The image classification is a CNN model. It feeds on the following dataset: https://www.kaggle.com/datasets/jessicali9530/stanford-dogs-dataset, and uses the PyTorch ML framework library with the ResNet18 image classifier pre-trained on ImageNet images.

Possible scenarios

This application targets city dwellers with pets who are constantly surrounded by other dogs. Our dog classification technology will help dog lovers identify breeds and match them with compatible furry friends.

PROFESSIONALS

Dog sitter: match your dogs

Mr. X is a professional dog-sitter and lives comfortably from his activity. One of his clients, Mrs. Y owns a Chihuahua (70% probability). A new client comes along, but Mr. X is reluctant to sign a new contract. He uses the application to determine the breed of the dog (55% probability of being a Rottweiler, 30% probability of being a Labrador, 15% undefined). Will the genes of the Rottweiler match the loquacious character of the Chihuahua?

ADOPTION

Learn more about your new adoption

Mr. X. is in the market to adopt a dog but would love to take home one of the many shelter dogs looking for a home. Knowing the personality and tendencies of his new dog is important as he trains his new puppy. This application will be perfect as Mr. X chooses his new furry friend.

NEW PALS

Meet your dog's soulmate

Mr. X is a very busy man but loves his dog! Since he is not often around, he wants his dog walks to be qualitative. For this, he uses the application to find the ideal playmate (criteria based on the image of each dog breed). Will he be a great athlete like the German Shepherd or more of an affectionate type like the Labrador?