Inventorious --- A Real-Time Grocery Shelf Manager

The Project

To tackle the problem of food waste, I aimed to help the suppliers, one of the main contributors, by increasing the transparency of grocery inventory. Therefore, I created this mobile application as a platform where consumers can share real-time updates of shelf stocks for others, who can then plan their shopping and obtain items from other sources. More importantly, supermarket managers can have a better idea of the demand for different products at their location and accordingly adjust their imports. With these effects, the app promotes utilization of food resources and minimizes products that serve no use. The app is composed of two main pages, a store pages, where users can navigate through a store for food items, and a wishlist page, where suppliers can plan out their imports and consumers can plan out their shopping. A key feature of this app is the implementation of computer vision, which can be activated through the camera. As a user takes a photo of the shelf, the model identifies and counts the items in the photo for the user to update supermarket stock information with. The project used Dart, a programming language, Roboflow, YoloV5, and Firebase in the development process.

Mobile
Community
Environment

Team Comments

I chose to make this project because...

Food waste is arguably the greatest issue in the world, and 13% of it comes from retail. It's an economically and environmentally tolling issue. One of the causes could be suppliers' uncertainty of demand, leading to plethora that's eventually wasted. Thus, I aim to increase transparency for both.

What I found difficult and how I worked it out

Designing the layout and function of the app with user-freindliness, which makes people beware of the the app's purpose. Accomodating a variety of grocery items in the dataset while maintaining model accuracy. Creating a secure server for the app's real-time nature.

Next time, I would...

I would refine my machine learning algorithm even more, by either implementing new research results on computer vision or training the model with more diverse pictures and conditions.

About the team

  • Canada

Team members

  • Tianyu (Lawrence)