Encoded Deep Features for Visual Place Recognition

Abstract

In this work, a new VPR approach that uses the features extracted from a Convolutional Neural Network (CNN) architecture that will be encoded by the Fisher Vector (FV) is introduced. As the main aim of this work is to develop a robust approach that can meet real-life challenges, the deep features are encoded with FV, which as shown in the experiments section, can lead to getting more robust features. Our approach was evaluated using two classifiers, Dynamic Time Warping (DTW) and Support Vector Machine (SVM) in particular. Using both classifiers, the FV-based encoded features have outperformed the non-encoded features.