Pocket Pantry

• 

  Pocket Pantry!

Inspiration

Environmental Conservation!

What it does

• Seamless shopping list experience.
• Track your pantries expiration dates using AI.
• Alert you when things are about to go bad.
• Use soon-to-expire ingredients to generate recipes.

How we built it

• We used Next.js to simulate a mobile application, as no one on the team had any mobile development experience. Further development would involve translating it into _ react native _ or _ flutter _ for use as a native app on mobile devices.
• We use FlaskAPI to communicate data between our front end and Firebase data storage. It's also the medium sends and receives data to our two different models.
• We utilized _ Huggingface's _ Distilroberta-base, fine-tuned using transfer learning techniques to fill-mask and esimate the expiration date based on the food and environment it's stored.
• We used OpenAI's GPT-3.5 to generate recipes based on the soon-to-expire ingredients, supplemented by whatever you have on hand.

Challenges we ran into

No one on our team has mobile app development. We played through our strengths and developed it using mobile frameworks, designed to be used on your phone. In the future, this will let users use our software without even downloading the app! Further, it is quite difficult to train a deep learning model on a laptop. Ouch.

Accomplishments that we're proud of

We have two first time hackathoners! I (John) am very proud of Zachary and Matthew for completing their first time hackathon!

What we learned

How to make a fullstack app! Transfer Learning a pre-trained AI model! Coding!!!!

What's next for Pocket Pantry

• Translate to native phone using react-native.
• Add user authentication and modify database to handle.
• Fix bugs.
• Deploy to your nearest app store!

Built With

• deep-learning
• flask
• huggingface
• javascript
• machine-learning
• nextjs
• python