Skip to content

A nonlinear classifier for categorizing shoes using machine learning

License

Notifications You must be signed in to change notification settings

zakwht/sneaker-or-sandal

Repository files navigation

Sneaker or Sandal SVM Classifier

License Accuracy

A classification model that uses a machine learning to categorize an image of a shoe as either a sneaker or a sandal. The app converts images into a 2D array of pixels (28x28) which are classified by a support vector machine (SVM) powered by a Gaussian (RBF) kernel.

  • The model's parameters are the regularization parameter C = 4 and the kernel coefficient γ = 0.036. These were the results of hyperparameter tuning with seven-fold cross-validation.
  • The training set had 4000 images, and the test set had 667 (6:1 split). The model ultimately performed with 98.4% accuracy based on the data.
  • The 95% confidence interval for the test error is (0, 0.0687).

Demo

Development

Requirements are cataloged in requirements.txt.

The classifier can be run from the command line with make, which will run the site from localhost:5000.

Acknowledgments