Sketchify: A Drawing Classification Toolkit Based on the Google Quick Draw Dataset

Sketchify is a drawing classification toolkit based on the Google Quick Draw dataset, implementing various classifiers from traditional machine learning (e.g., Naive Bayes, KNN, SVM, XGBoost) to deep learning (RNN). It is an excellent practical case for learning image/sequence classification algorithms.

• Original Author/Maintainer: slxppin
• Source Platform: GitHub
• Original Link: https://github.com/slxppin/Sketchify-A-Quick-Draw-drawing-classifier
• Release Date: 2026-06-07

The Google Quick Draw dataset used in the project is a large-scale hand-drawn sketch dataset released by Google, containing millions of simple graphics drawn by users. Each sample is a sequence of stroke coordinates, suitable for sequence modeling.

Classifier Implementation

• Traditional Machine Learning: Gaussian Naive Bayes, KNN, SVM, Logistic Regression, XGBoost
• Deep Learning: PyTorch-based RNN (captures stroke temporal dependencies)

Feature Engineering

• PCA dimensionality reduction
• Sequential Forward Selection (SFS) feature selection
• K-fold cross-validation to ensure robust evaluation

Tech Stack

Python, NumPy, Pandas, Matplotlib, Seaborn, Scikit-learn, PyTorch, XGBoost

Project Structure

Each classifier has an independent script (e.g., gaussian_naive_bayes.py, rnn.py), and the visualization script is visualize_results.py

Data Visualization

Implemented using Matplotlib and Seaborn: confusion matrix, feature importance, PCA visualization

Usage Workflow

1. Clone the repository
2. Install dependencies
3. Download the Google Quick Draw dataset and place it in the data directory
4. Run the classifier scripts
5. Execute the visualization script

Learning Value

• Compare multiple algorithms on the same dataset
• Practice feature engineering techniques
• Get started with sequence modeling (RNN application)
• Demonstrate a complete end-to-end workflow

Extension Possibilities

• Try Transformer, CNN+LSTM hybrid architecture
• Implement a real-time drawing recognition interface
• Add data augmentation
• Multi-label classification system
• Mobile application

Summary

Sketchify has a clear structure and complete documentation, demonstrating best practices in classification algorithms, feature engineering, and visualization. It is an excellent case for learning image/sequence classification.