Live Website: https://shazam4petz.github.io/
Create a high-accuracy tool for shelters to correctly identify breed combinations and percentages of cross and mixed-bred dogs and cats using machine learning, and help raise adoption rates of shelter animals as a result.
According to an article in the Smithsonian (2018), roughly 3.3 million dogs enter animal shelters in the U.S. each year, but only an alarming 67% are correctly identified of their primary or secondary breed by shelter staffs. Leaving over 1.089 million dogs across the country to be misidentified or mislabeled yearly.
Currently, there are no standardized procedures across the U.S. for identifying breeds of dogs and cats. The accuracy rate for breed-identification varies significantly from shelter to shelter and staff to staff. Studies show that as high as 48% of dogs with no genetic evidence of being Pit Bulls have been misidentified as Pit Bull-type dogs. Widely (and incorrectly) believed to be "dangerous" and aggressive in nature, Pit Bulls have one of the lowest adoption rates, and maintains to be the most commonly restricted dog breed from "pet-friendly" rental properties.
Inaccurate breed-labeling of shelter animals drastically limits their pool of potential adopters, and as a result, directly lowers their chances of being adopted.
| 🏆 Best Use of Google Cloud Platform | 🏆 Best Almost Hack |
|---|---|
 |
 |
- Firebase: Cloud Storage, Realtime Database, User Authentication
- Google Cloud Vision API: Image Analysis
- Google Cloud AutoML Vision: Custom Machine Learning Model Training
- Google Cloud Functions: Serverless Application Scripts
- Google Colab: Collaborative Python IDE
- Further train our Model with more data, improving its precision as it scales
- Implement a validation/filtering algorithm for photos uploaded by users
- Store approved photos uploaded by users to our database for further Model-training
- iOS & Android Apps: allowing photos taken via mobile phones to be analyzed direct within our app
- Using the Wolfram Alpha API, provide users with health information for specific mix-breeds of dogs and cats
- Expand our coverage to other species of animals
Brought together by our mutual love for animals, we found each other at the team-forming workshop of Hack(H)er413, a 2-day hackathon from Feb 09-10, 2019, on the UMass Amherst campus.
- Miffy Chen: Website, Google Cloud Platform, Design, Integration
- Anna Mun: Cloud Vision API
- Qing Zhao: AutoML Model
- Stephanie Murphy: REST API