Skip to content

Used machine learning concepts in Python to predict the test-set for CIFAR-10 with approx. 70% accuracy

Notifications You must be signed in to change notification settings

aatmiyasilwal/CIFAR-10-Classification

Repository files navigation

COMP3314 Image Classification Kaggle Competition - Team ADventurers

Team members: Aatmiya Silwal, Rishi Shah, Sai Sinn Zom Leng

This project makes use of the CIFAR-10 dataset (https://www.cs.toronto.edu/~kriz/cifar.html) and uses non-neural network-based classifiers to predict the labels of the test set.

By making use of feature extraction methods like Scale-Invariant Feature Transform (SIFT), Histogram of Oriented Gradients (HOG), Color histograms, and Edge Orientation Histograms, we were able to achieve a final accuracy score of 69.8%.

To further expand the training dataset, image augmentation was utilized, which was followed by the use of RBF Support Vector Machine (after dimensionality reduction) to predict the class labels.

About

Used machine learning concepts in Python to predict the test-set for CIFAR-10 with approx. 70% accuracy

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published