Learning to Assign Orientations to Feature Points

TL;DR

This paper introduces a neural network-based method to assign canonical orientations to feature points, improving matching performance across various datasets without dataset-specific retraining.

Contribution

It presents a novel Siamese network approach with a new activation function to learn orientations, enhancing feature matching accuracy.

Findings

Outperforms state-of-the-art methods on multiple datasets

Works with any rotation-sensitive descriptors

Eliminates need for retraining on new datasets

Abstract

We show how to train a Convolutional Neural Network to assign a canonical orientation to feature points given an image patch centered on the feature point. Our method improves feature point matching upon the state-of-the art and can be used in conjunction with any existing rotation sensitive descriptors. To avoid the tedious and almost impossible task of finding a target orientation to learn, we propose to use Siamese networks which implicitly find the optimal orientations during training. We also propose a new type of activation function for Neural Networks that generalizes the popular ReLU, maxout, and PReLU activation functions. This novel activation performs better for our task. We validate the effectiveness of our method extensively with four existing datasets, including two non-planar datasets, as well as our own dataset. We show that we outperform the state-of-the-art without the need of retraining for each dataset.