Randomized Ferns on Android

Advisor: Prof. Dimitris Samaras

This project is about recognizing the patches surrounding key-points, our classifier uses hundreds of simple binary features and models class posterior probabilities. We make the problem computationally tractable by assuming independence between arbitrary sets of features. Even though this is not strictly true, we demonstrate that our classifier nevertheless performs remarkably well on image datasets containing very significant perspective changes. We have developed an Android cell phone application wherein we will be training randomized ferns with several images of the buildings. This information was then used to retrieve relevant information from the classifier from each frame of the user’s mobile device. This report contains all the relevant implementation details and our results. Developed an Android application for building recognition using randomised ferns, a semi Naive Bayes classification approach.

(Detailed Report)

#video