Fashion MNIST Classification Using Convolutional Neural Networks (CNN)

Overview

This project demonstrates a Deep Learning approach to classifying clothing items using a Convolutional Neural Network (CNN) trained on the Fashion MNIST dataset. By leveraging advanced neural network techniques, we've developed a model capable of accurately identifying various clothing categories with high precision.

The primary objective was to build a robust image classification model that can distinguish between 10 different types of clothing items. Our model achieved a training accuracy of 0.91 and a test accuracy of 0.88, showcasing its effectiveness in learning and generalizing clothing item features.

Dataset

Dataset: Fashion MNIST (TensorFlow/Keras Datasets)

The Fashion MNIST dataset is a comprehensive collection of grayscale images representing 10 different clothing categories. Each image is a 28x28 pixel grayscale representation of a clothing item, making it an ideal benchmark for machine learning image classification tasks.

Dataset Characteristics

• Total Images: 70,000
• Training Set: 60,000 images
• Test Set: 10,000 images
• Image Dimensions: 28x28 pixels
• Color: Grayscale
• Categories: 10 clothing types

Model Architecture

Our Convolutional Neural Network (CNN) architecture is designed to effectively extract and learn features from clothing item images:

Layer Configuration

1. Convolutional Layer 1: 30 filters, 3x3 kernel, ReLU activation
2. Max Pooling Layer 1: 2x2 pool size
3. Convolutional Layer 2: 60 filters, 3x3 kernel, ReLU activation
4. Max Pooling Layer 2: 2x2 pool size
5. Flatten Layer: Converts 2D feature maps to 1D feature vector
6. Dense Layer 1: 60 neurons, ReLU activation
7. Output Layer: 10 neurons (one per clothing category), Softmax activation

Training Configuration

• Optimizer: Adam
• Loss Function: Sparse Categorical Cross-Entropy
• Metrics: Accuracy

Results

Performance Metrics

• Training Accuracy: 0.91
• Test Accuracy: 0.88

The model demonstrates strong performance in classifying clothing items, with a slight variance between training and test accuracy indicating good generalization.

Dependencies

Required Libraries

• Python 3.8+
• TensorFlow 2.x
• Keras
• NumPy
• Matplotlib

How to Run

Using Google Colab

1. Open Google Colab:

   • Navigate to Google Colab
   • Click on "New Notebook" or "File" > "New notebook"

2. Set Up Runtime:

   • Go to Runtime > Change runtime type
   • Select Python 3 as the runtime
   • (Optional) Select GPU for faster training

3. Install Dependencies:

   !pip install tensorflow numpy matplotlib

4. Clone the Repository (if using GitHub):

   !git clone https://github.com/Nirikshan95/Fashion-MNIST-Classification-Using-Deep-Learning.git

5. Open the Notebook:

   • Upload or open the Jupyter notebook for the project
   • Run cells sequentially
   • The dataset will be automatically downloaded via TensorFlow/Keras

Local Development Alternative

If you prefer local development:

1. Clone the repository:

   git clone https://github.com/Nirikshan95/Fashion-MNIST-Classification-Using-Deep-Learning.git
   cd Fashion-MNIST-Classification-Using-Deep-Learning

2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Run the Jupyter Notebook or Python scripts as needed

Repository Structure

Fashion-MNIST-Classification-Using-Deep-Learning/
│
├── fashion_mnist_classification.ipynb             # notebook
├── requirements.txt       # Project dependencies
└── README.md              # Project documentation

Future Work

• Implement data augmentation techniques
• Experiment with more complex CNN architectures
• Explore transfer learning approaches
• Add real-world clothing item classification support

License

Dataset

The Fashion MNIST dataset used in this project is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0).

Contact

For questions or collaboration, please contact:

• Your Name
• Email: nirikshan987654321@gmail.com
• GitHub: @Nirikshan95