Doodle-Classifer!

This project implements a machine learning model to classify hand-drawn sketches using the Quick, Draw! dataset. It includes data preprocessing, model training, evaluation, and a GUI application for testing the model.

Project Structure

quickdraw_project/
│
├── quickdraw_data/         # Raw .ndjson files from Quick, Draw! dataset
├── datasets/               # Processed datasets
├── models/                 # Saved models and visualizations
├── scripts/               
│   ├── preprocess_data.py  # Data preprocessing
│   ├── train_model.py      # Model training
│   ├── evaluate_model.py   # Model evaluation
│   └── gui_app.py          # GUI application
├── requirements.txt        # Dependencies
└── README.md              # This file

Setup and Installation

1. Create a virtual environment (recommended):

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

2. Install dependencies:

   pip install -r requirements.txt

3. Download Quick, Draw! dataset files (.ndjson format) and place them in the quickdraw_data folder.

Usage

1. Preprocess the data:

   python3 scripts/preprocess_data.py

2. Train the model:

   python3 scripts/train_model.py

3. Evaluate the model:

   python3 scripts/evaluate_model.py

4. Run the GUI application:

   python3 scripts/gui_app.py

Model Architecture

The model uses a CNN architecture with:

• Multiple convolutional layers with batch normalization
• Max pooling layers
• Dropout for regularization
• Dense layers for classification

Performance

The model's performance metrics and visualizations will be saved in the models directory after evaluation:

• confusion_matrix.png: Visualization of model predictions
• prediction_samples.png: Sample predictions on validation data
• training_history.png: Training and validation metrics over time

GUI Application

The GUI application provides a simple interface to:

• Draw sketches
• Get real-time predictions
• Clear the canvas
• View prediction confidence

Dependencies

• TensorFlow
• OpenCV
• NumPy
• PyQt5
• scikit-learn
• matplotlib
• tqdm

Source

The data used in this project is sourced from the Quick, Draw! dataset.