BergNet: Classifying Harvard Dining Hall Food Using Deep Learning

Overview

BergNet is a deep learning project which classifies food items served at Harvard University’s Annenberg Hall. This project uses a custom, original dataset of images taken from two days at Annenberg Hall and builds neural networks off of it to analyze and recognize a variety of meals, enhancing meal recommendations and food monitoring in the dining industry.

Project Structure

The project folder contains the following components:

• models/: This directory contains the trained models and related files used for food classification, in Mathematica object format.

  • trainedEfficientNet.mx: The trained EfficientNet model.
  • trainedBergNetNoAug.mx: The trained BergNet model without data augmentation.
  • trainedBergNet.mx: The trained BergNet model with data augmentation.

• data/: This folder includes the datasets used for training and testing the models.

  • train-set.mx: The training dataset used to train the models.
  • test-set.mx: The testing dataset used to evaluate the models.
  • augmented-train-set.mx: The training dataset that includes augmented data for better model generalization.
  • raw-data-wednesday/: This folder contains the raw food images and labels collected on Wednesday.
  • raw-data-thursday/: This folder contains the raw food images and labels collected on Thursday.

• BergNet-Report.pdf: A comprehensive report detailing the project methodology, results, and findings.

• BergNet-Presentation.pdf: A presentation summarizing the project.

• BergNet-MathematicaCode.nb: A Mathematica notebook containing the code used for training and evaluating the models.

Usage

To get started with the project, follow these steps:

1. Clone the Repository:

   git clone https://github.com/lindazeng979/BergNet.git
   cd BergNet

2. View Reports and Presentations: Open the BergNet-Report.pdf and BergNet-Presentation.pdf using any PDF viewer to explore the detailed insights into the project and its findings.

3. Run Mathematica Code:

   • Ensure you have Mathematica installed on your computer.
   • Open the BergNet-MathematicaCode.nb notebook in Mathematica.
   • Ensure that the notebook is in the same directory as the models and data folders.
   • To run the code, click on the "Evaluate" menu and choose "Evaluate Notebook" or use the keyboard shortcut Shift + Enter for each selected cell.
   • You can also modify the code and parameters as needed to experiment with the model training and evaluation.

Contact

For any questions or inquiries, please contact: Linda Zeng
Email: 26lindaz@students.harker.org

Acknowledgments

Thanks to Harvard University's Pre-College Program for providing the opportunity to explore this project.