Inspiration
We spoke extensively with members of the Gate Group team to understand the main challenges they faced and how intelligent automation could be integrated into their production process without disrupting existing structures. Our goal was to design a system that enhances quality control while remaining simple and intuitive for workers to use.
What it does
Our solution uses a computer vision model to automatically detect errors during the packaging process and report them in real time. It also collects and stores these insights to enable supervisors to monitor performance and better understand operational metrics.
How we built it
We gathered hundreds of photos of real-life meal trays and uploaded them to Roboflow, where we manually tagged each item. These labeled datasets were then used to train a YOLOv8 computer vision model capable of identifying and counting tray components. The system sends instant alerts to workers when discrepancies are detected, allowing them to correct issues before the product reaches the client.
Assumptions
- Supervisors record and report the number of errors made by employees.
- The system tracks the inflow and outflow of food items to ensure accurate inventory and production balance.
- Cameras are placed in consistent, well-lit positions to capture reliable data for detection and analysis.
Impact
By automating error detection and data collection, our solution reduces human error, minimizes waste, and increases client satisfaction. At the same time, it empowers supervisors with real-time insights and performance metrics to improve overall efficiency and decision-making.
Challenges & Learnings
One of the main challenges was dealing with lighting variations and dataset consistency. Through testing and iteration, we learned to fine-tune the YOLOv8 model and enhance labeling accuracy to achieve better precision.
Log in or sign up for Devpost to join the conversation.