Welcome to our GlucoScan project! This repository houses a powerful machine learning solution designed to analyze nutrition facts on drink labels. Using image recognition and Optical Character Recognition (OCR), we identifies and extracts critical details — specifically the sugar content — making it easier to stay informed about your drink choices.
- Image Recognition: Detect and locate nutrition fact tables on drink labels.
- OCR Technology: Extract and interpret text from the detected tables with high accuracy.
- Focus on Health: Automatically capture the sugar amount to help users make healthier choices.
- Upload an image of a drink's nutrition facts label.
- The ML model detects the table containing nutritional details.
- OCR extracts the relevant data, focusing on the sugar content.
- Output: The sugar amount in grams is displayed or stored for further use.
Here's what you'll find in this repository:
- Image_Recognition_+_OCR.ipynb: Interactive Jupyter Notebook for testing and experimenting with the model.
- image_recognition_+_ocr.py: Python script to run the full pipeline outside of the notebook environment.
- final_nutrition_model_ML.h5: Our machine learning model used for image recognition and OCR tasks.
- requirements.txt: A list of dependencies to install for running the project.
Clone the repository: git clone https://github.com/GlucoScan-Bangkit/GlucoScanProject.git
cd GlucoScanProject
Install dependencies: pip install -r requirements.txt
Run the pipeline: Notebook: Open Image_Recognition_+OCR.ipynb in Jupyter Notebook to interactively test the system. Script: Run the pipeline using the Python script: python image_recognition+_ocr.py
You can access our dataset from: https://drive.google.com/drive/folders/1Casu1_jepy_A4iI1gde8J5Z9Faz5faW8?usp=drive_link
We welcome contributions! Feel free to open an issue or submit a pull request to enhance the project.