Find My Pet

Welcome Screen
Sign In Page (the white corners are not actually like that)
Comparing two posts
User Profile Page

Inspiration

Around my neighborhood and town, I've noticed a lot of missing pets. The only "solution" I've seen is posting flyers around and hoping that someone pays attention and looks out for the pet on the flyer. However, this is extremely ineffective. I've gone through losing a pet and I would hate for someone to go through the same because of a lack of a good solution.

What it does

Find My Pet allows users to create posts of pets that they have lost, or of roaming pets that they have seen/found. Upon logging in, users are greeted with a dashboard containing all of the posts that users have created. When creating a post, users are able to fill in details like what species the pet is, where they last saw the animal, and when they last saw it. In order to create a post, users must also input an image of the pet for better identification. This image can be compared to images from other posts within a given area to be passed into a ML model to check for similarity.

How I built it

For user authentication, I used Firebase. In order to store user data along with posts, I used Firestore. I utilized CreateML to create my model. The model was trained on the Stanford Pet Dataset for 25 epochs and achieved an accuracy of 75%. For the UI, I utilized SwiftUI.

Challenges we ran into

I was heavily pressed for time this Hackathon as I spent a lot of time trying to import a sample TensorFlow model into my XCode project, which never worked. I also had a lot of nuance errors when setting up Firestore, which was much more difficult than I thought it would be. I also didn't have time to actually create or implement a customized model created in Python, so I had to use CreateML instead, which achieved a lower accuracy score than a custom model probably would have.

Accomplishments that we're proud of

I'm proud of my use of Firestore and Firebase for user authentication and data storing. For a model that I didn't customize, I'm happy with an accuracy of 75%. I'm also proud of the sign-in UI.

What we learned

I partially learned how to import a TensorFlow model into an XCode Project. I learned how to better utilize Firestore, and how to create cleaner-looking UIs. I also learned not to worry about things that don't matter as much and to focus on the main aspects of my project first.

What's next for Find My Pet

I hope to implement my own ML model in order to increase accuracy. I also hope to tweak the UI to be consistent throughout the entire app. Adding more support for user-profiles and interaction on posts is also a high priority. When I've added all of the features that are necessary, I hope to bring it to the App Store.