Download attachment GitHub repository
The objective of this project is image classification using kernel methods. Surprisingly good performances are obtained training a kernel ridge regression model for classification with intensive data augmentation. Image features are extracted using histograms of oriented gradients.
The method allowed to reach the 3rd position of the data challenge leaderboard, on both public and private data.
This project was carried in 2021/22 as part of the data challenge for the Kernel Methods in Machine Learning course, for the master’s degree in Data Science and Artificial Intelligence at Grenoble INP - Ensimag.