Due to my negligence and lack of time management, I was only able to start my FYP proper this week. I was briefed on the idea of this FYP about using AR for military training purposes, namely the CBRE.
I was tasked to read up on the Ferns algorithm. According to “Fast Keypoint Recognition usng Random Ferns”, it is an algorithm that finds the most stable points of an image through classification of patches of the image. Naive Bayesian Classification is used here, as well as k-means clustering to partition the classifications of the patches in order to classify the patches. Next, a series of tests are done on the patches, which is where the difference between randomized trees and ferns lies. Randomized tree executes a hierarchy of tests onto the patches, while fern executes a series of tests, one test after another, onto the patches. Hence fern is much less computational intensive than randomized tree.