VidLoc: A Deep Spatio-Temporal Model for 6-DoF Video-Clip Relocalization

TL;DR

This paper introduces VidLoc, a deep recurrent model that leverages temporal information in video sequences to improve 6-DoF camera localization accuracy and smoothness, outperforming single-image methods.

Contribution

The paper presents a novel recurrent neural network approach for 6-DoF video localization that exploits temporal smoothness, reducing localization errors significantly.

Findings

Short sequences (20 frames) improve pose smoothness and accuracy.

The model outperforms single-image localization methods.

Probabilistic pose estimates are feasible with the proposed approach.

Abstract

Machine learning techniques, namely convolutional neural networks (CNN) and regression forests, have recently shown great promise in performing 6-DoF localization of monocular images. However, in most cases image-sequences, rather only single images, are readily available. To this extent, none of the proposed learning-based approaches exploit the valuable constraint of temporal smoothness, often leading to situations where the per-frame error is larger than the camera motion. In this paper we propose a recurrent model for performing 6-DoF localization of video-clips. We find that, even by considering only short sequences (20 frames), the pose estimates are smoothed and the localization error can be drastically reduced. Finally, we consider means of obtaining probabilistic pose estimates from our model. We evaluate our method on openly-available real-world autonomous driving and indoor localization datasets.