🧥 Fashion Classifier (MNIST) – TensorFlow, Flask, and FastAPI

This project is a full-stack web application that allows users to upload an image of a fashion item (e.g., shirt, sneaker, bag) and classifies it using a Neural Network (NN) and Convolutional Neural Network (CNN) trained on the Fashion Classifier dataset.

The app features:

• 🔍 Real-time image classification
• ⚙️ NN and CNN model build separately
• 🧠 Robust preprocessing (resize, normalize, center the object)
• 📸 Live preview of uploaded image
• 💡 Displays top 3 predictions with confidence scores
• 🎨 Beautiful and responsive HTML/CSS interface

🚀 Live Demo

Want to see it in action?
👉 [Coming Soon: Deployed on Render or Hugging Face Spaces]

🧠 Model Details

• Architecture: 3-layer CNN with BatchNorm, Dropout, and MaxPooling
• Input shape: 28×28 grayscale images
• Training Dataset: Fashion MNIST (60,000 training + 10,000 test images)
• Accuracy: ~91% on test data

---

🖼️ Supported Classes

['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat','Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']

---

🧪 Try It Locally

1️⃣ Clone the repo

git clone https://github.com/Aronno1920/Fashion-Classifier-MNIST.git
cd fashion-classifier-mnist

2️⃣ Install requirements

pip install -r requirements.txt

3️⃣ Run the app

python app.py

Open your browser: http://127.0.0.1:5000

📁 Project Structure

├── ImageSample/                        # Folder for example images or datasets
├── __pycache__/                        # Python cache files (auto-generated)
├── model/
│   ├── best_nn_model.keras             # Saved trained NN model weights
│   └── best_cnn_model.keras            # Saved trained CNN model weights
├── notebooks/
│   ├── fashion_mnist_selim_ahmed.py    # Saved colab file as python
│   └── Fashion_MNIST_Selim_Ahmed.ipynb # Saved colab file 
├── static/                             # logo, style and other necessary
│   ├── logo.png
│   └── style.css
├── templates/                          # Templates for web app or reports (if any)
│   ├── index.html
│   └── train.html
├── templates/                 
├── README.md                           # Project description and documentation
├── app.py                              # Main application script (e.g., for running or serving the model)
├── model_config.py                     
├── train_model_cnn.py                  # CNN model training class
└── train_model_nn.py                   # NN model training class

🛠 Tech Stack

    Python
    TensorFlow / Keras
    Flask
    PIL (Pillow)
    HTML5 + CSS
    JavaScript (Image preview)
    SciPy (for center-of-mass preprocessing)

📸 Screenshots

✍️ Author

Selim Ahmed
📫 Connect on LinkedIn
🌐 GitHub Profile