README for DS340-CNN-Project

DS340 CNN Project: Waste Classification Using Convolutional Neural Networks (WasteWise AI)

This project focuses on building and fine-tuning a Convolutional Neural Network (CNN) to classify waste into three categories: Food Waste, Mixed Recycling, and Trash. The ultimate goal is to improve waste sorting systems and promote sustainability by leveraging machine learning. The repository includes the base model, experiments with data augmentation, fine-tuning techniques, and a prototype design for an intelligent trash disposal station.

Project Structure

DS340-CNN-Project/
│
├── waste-images/                           # Directory for training, validation, and test data
│   ├── train/                     # Training dataset
│   ├── val/                       # Validation dataset
│   └── test/                      # Testing dataset
│
│
├── notebooks/                     # Jupyter notebooks
│   └── GAN.ipynb                  # Exploratory Data Analysis
│   └── Experiments.ipynb    # Model building and fine-tuning
│   └── synthetic_data.ipynb
├── model_predictions.html         # Interactive HTML report of predictions
└── README.md                      # Project description and instructions

How to Run the Model

Prerequisites

Ensure you have the following installed on your system:

• Python 3.7 or later
• TensorFlow 2.x
• Jupyter Notebook (optional, for running .ipynb files)
• Required Python libraries (install using requirements.txt)

Steps to Run

1. Clone the Repository

   git clone https://github.com/Louie-Belile/DS340-CNN-Project.git
   cd DS340-CNN-Project
   

2. Prepare the Data Ensure the dataset is structured as follows:

   data/
   ├── train/
   │   ├── food_waste/
   │   ├── mixed_recycling/
   │   └── trash/
   ├── val/
   │   ├── food_waste/
   │   ├── mixed_recycling/
   │   └── trash/
   └── test/
       ├── food_waste/
       ├── mixed_recycling/
       └── trash/
   

3. Run the Base Model Use the main.py script to train and evaluate the base model:

   python main.py
   

4. Explore Fine-Tuning Experiments Open the model_fine_tuning.ipynb notebook in Jupyter Notebook to explore experiments on hyperparameter tuning, data augmentation, and architecture adjustments:

   jupyter notebook notebooks/model_fine_tuning.ipynb
   

5. Evaluate the Model Evaluate the trained models on the test set using the provided scripts or the interactive model_predictions.html file.

6. View Results Visualizations and metrics such as confusion matrices, training history, and predictions can be found in the results/ folder.

Key Features

• Base Model: A CNN trained with three convolutional layers and dropout regularization.
• Data Augmentation: Techniques like rotation, zoom, and brightness adjustment to improve generalization.
• Fine-Tuning: Experiments with dropout rates, learning rates, and model architecture adjustments.
• Interactive Results: View predictions and their confidence levels in the model_predictions.html file.
• Prototype Design: Intelligent trash disposal system prototype included.

Notes

• The model achieves promising classification results, particularly in the underrepresented category of food waste, demonstrating the importance of addressing data imbalance.
• Fine-tuning experiments highlight potential avenues for improving performance in future iterations.
• Compressed Data can be found at this link -> https://drive.google.com/file/d/1mAlNb0658I-SeSv-pe9998w7bVPKi68u/view?usp=sharing