Grocery Detection API using YOLOv8

Overview

This project is a FastAPI-based application for real-time grocery item detection using a custom-trained YOLOv8 model. It identifies grocery items such as apples, tomatoes, bananas, and grapes from uploaded images.

Features

• Custom YOLOv8 model for grocery detection.
• Supports real-time prediction with confidence thresholding.
• Outputs predictions with labels and confidence scores.
• API support for client applications.

---

Requirements

• Python 3.8+
• ultralytics==8.0.0 (or later)
• OpenCV
• FastAPI
• Uvicorn

Install Dependencies

pip install ultralytics opencv-python fastapi uvicorn

---

Running the Server

1. Start the FastAPI server:

uvicorn main:app --host 0.0.0.0 --port 8000 --reload

2. Use ngrok to expose the server to the internet:

ngrok http 8000

3. Copy the public URL provided by ngrok and use it to access the API endpoint:

POST {ngrok-url}/predict/

---

Results

• Outputs detected objects with confidence levels.
• Displays label with the highest confidence.

---

Notes

• Ensure the dataset is properly labeled and organized.
• Adjust confidence threshold (conf=0.6) as needed for performance.
• Update best.pt with the latest trained model weights.

---

License

This project is licensed under the MIT License.