HodgePodge: Minimize Waste, Maximize Flavor!

Demonstration of image transformations applied before image is fed into OCR CV model.
Sample chat log with inventory chat bot.

Food for thought: Every year, 80 million tons of food waste worth $444,000,000,000 ($444 billion) are tossed into the mish mash of America’s landfills (source: Feeding America). That’s not even including the copious amounts of land, water, time used, and environmental damages inflicted from producing all of these food. Restaurants in particular, are a key ingredient as the root of this problem - they generate roughly 20 to 33 billion pounds of food waste each year. A plethora of solutions regarding this problem have been proposed that involve rerouting food that isn’t used by restaurants to those who need them. However, coordinating such distribution across restaurants, the hungry, the government, and shifting the collective toward a completely new paradigm is a logistical nightmare. Instead of taking reactionary measures, lettuce (let us) fight the food waste problem at its root: overproduction. Specifically, since 99% of American restaurants are family-owned businesses that employ fewer than 50 people, many can’t afford to cook up a massive system-wide inventory manager. Despite this, they still need to deal with the culinary chaos of a bustling restaurant kitchen, where every expiration date is a ticking time bomb. Yet, frequently restocking ingredients also introduces immense overproduction, leading to food waste and profit losses. From these motivations, our team has stirred up the concept of HodgePodge, an easily accessible app that makes managing kitchen inventories easier than making a grilled cheese, and even more satisfying than biting into one. With this vision, HodgePodge simmers efficiency and sustainability together into a secret sauce that's both practical and savory!

What it does: HodgePodge is a Swift, sleek inventory management app designed specifically for smaller, family owned restaurants that want to minimize waste and maximize flavor. With an encrypted SQLite3 backend tougher than a well done steak, your food inventory is secure, organized, and easily accessible. Adding new items is as easy as pie (and just as satisfying). Our app recognizes names of items, quantities, purchase dates, and expiration dates, ensuring your inventory is as fresh as your dishes. Need to reduce or delete items? It's just a tap away. But that's not all - we've introduced a state-of-the-art cutting-edge CV model that scans expiration dates from quick snapshots of food items. Say goodbye to manual entry and hello to efficiency! But wait, there's more! Don’t you wish sometimes that you could just ask your inventory what’s in stock instead of going through countless printed logs? Well, now you can. Our chat feature allows you to ask questions about your current stock in LLM. With our integrated chat feature, ask any question about your inventory in natural language and receive smart, analytics-driven feedback. Wondering if you have enough tomatoes for tonight's service or if the fish should be used by tomorrow? Ask away, and receive intelligent, analytics-backed feedback. It's like having a sous chef who majored in math. Whatever inventory-related questions you have, HodgePodge will deliver the answers to you spick and span, helping you make informed decisions that minimize waste and maximize flavor.

How We Built It: Brewed in the Treehacks hackathon cauldron, HodgePodge blends the robustness of SQLite3 encryption with the flair of SwiftUI for a seamless, secure frontend experience that is also a feast for the eyes. We applied various image transformations to the expiration date images to convert them from hard-to-interpret dot matrix form to smooth letters, which were fed to our CV model. Our model employs the fast, efficient and accurate EasyOCR library to localize and output expiration dates. HodgePodge’s intuitive chat feature is powered by a custom GPT 3.5 with full knowledge of the inventory database. Managing so many diverse ingredients for such a complex recipe was no easy task, but with Flask, we fluidly integrated each of these components into the HodgePodge user interface.

Challenges We Ran Into: The development of HodgePodge was no piece of cake. Despite the cutting-edge nature of our CV model, it couldn’t properly process the raw photos of the dot-matrix expiration dates of food labels. We spent hours experimenting with different combinations of image transformations and processing techniques before settling on a configuration that helped our CV model truly take off. Figuring out the communication between our Python scripts and our Swift interface added a particularly feisty cook to the kitchen. Initially, trying to facilitate communication between Swift and Python was like trying to mix oil and water. We had all of the ingredients for an amazing soup, but no pot to cook them in, until we finally found our cooking vessel - Flask.

Accomplishments That We're Proud Of: Seeing HodgePodge go from a jumble of ideas we excitedly pasted onto a spreadsheet at the beginning of Treehacks to a harmonious and fully functioning prototype with efficient and robust features, we're particularly proud of how we delivered an app that integrates so many useful and diverse features to help a family-owned restaurant manage their food waste while providing much-needed assistance to family-owned restaurants. We’re also extremely happy to see a full implementation and showcase of our idea through a color-schemed app for iOS. Not only did we demonstrate our idea, we displayed it with brilliance.

What We Learned: Developing HodgePodge has been a culinary journey of discovery. We ventured into many territories of software engineering, data and computer science that we weren’t familiar with, like iOS app development, optical character recognition, building custom GPTs, and Flask, Python, and Swift integration. Blending technical skill with empathy, we also learned that the key ingredients to a successful app include not just technical skills but also a deep understanding of the user's needs. HodgePodge is all about ease of access, and we spent a great deal of time debating about how to make its features as intuitive and easy to use as possible, amassing crucial knowledge about UX and frontend design.

What's Next for HodgePodge: As we look to the future, we aim to integrate AI-driven predictions for order suggestions based on inventory levels, usage patterns, and upcoming expiration dates, and expand the analytics tab in the app to include dynamic daily, weekly and monthly visual representations and graphs of these important restaurant statistics. We also want to bolster our CV-based expiration date detector to work on an even broader range of surfaces that require difficult and complex processing, such as waxy plastic and aluminum bags.

Conclusion: With HodgePodge, we're cooking up a medley of a solution that will transform the way family-owned restaurants operate their inventories and aid the global fight against reducing food waste by helping to alleviate the massive environmental impact of the restaurant industry.