VidLoc: A Deep Spatio-Temporal Model for 6-DoF Video-Clip Relocalization
This addresses incremental improvement in localization accuracy for applications like autonomous driving and indoor navigation by leveraging temporal constraints.
The paper tackles the problem of 6-DoF video-clip relocalization by proposing a recurrent model that exploits temporal smoothness, reducing localization error drastically with short sequences of 20 frames.
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.