This project was inspired by some innovative restaurant payment methods i've seen around, which implement real-time object detection to determine the price to pay.
I tried to recreate a similar system, developed as a simple web application using Flask, where the backbone network is a YOLOv8 model, which i fine-tuned on food images.
The price is adjusted in real-time as the webcam detects one or more food items, and the price tag is updated accordingly.
Here is the list of food items that the system can detect along with their corresponding prices:
- Banana: $5
- Black Beans: $4
- Grilled Chicken Breast: $7
- Milk: $2
- Orange Juice: $3
- Pizza: $8
- Potato: $3
- Salad: $5
- Spaghetti: $10
- White Rice: $5
These food classes are just the ones contained in the 'Food Detection Image Dataset', that i used for fine-tuning the YOLO model. However, it is fully customizable for personal use if someone wants to create a personal food dataset.
- Logic: Capture real-time video feed from the webcam and detect food items.
- Realtime Streaming: Stream video feed to the web browser.
This system uses a webcam to capture real-time video feed, which is then processed by a YOLOv8 model fine-tuned on food images. The model detects food items with a confidence threshold of 0.5 and calculates the total price based on the detected items. For instance, a detected 'spaghetti' adds $10 to the total, while a 'orange juice' adds $3.
- Logic: Calculate the total price based on detected items.
- Detected Items: Update price in real-time as new items are detected.
Users can press "Confirm" to generate a bill containing the price of each item detected along with the total price. The "Menu" button allows users to visualize the food list along with the associated prices.
The web application streams the video feed to the front end with bounding boxes and labels showing the detected items and their confidence scores, and updates the total price dynamically.
This ensures that users can see the price in real-time as they place more items in front of the camera.
- Video Capture: Using OpenCV to capture video from the webcam.
- Model: YOLOv8 fine-tuned on food images.
- Web Application: Flask for creating the web interface and streaming the video feed.
- Real-Time Updates: Flask-SocketIO for updating the price dynamically based on detected items.
Watch the demo video below to see the Real-Time Food Detection System in action:
Actual setup during video recording:
- Clone the repository and navigate to the project directory.
- Install dependencies:
pip install -r requirements.txt - Run the application:
python app.py - Navigate to
http://localhost:5000/
(if you enjoyed this content please consider leaving a star ⭐)