Unsupervised Domain Adaptation for 6DOF Indoor Localization

Department of Mathematics and Computer Science, University of Catania, Italy

D. Di Mauro, A. Furnari, G. Signorello and G. M. Farinella

Visual Localization is gathering more and more attention in computer vision due to the spread of wearable cameras (e.g. smart glasses) and to the increase of general interest in autonomous vehicles and robots. Unfortunately, current localization algorithms rely on large amounts of labeled training data collected in the specific target environment in which the system needs to work. Data collection and labeling in this context is difficult and time-consuming. Moreover, the process has to be repeated when the system is adapted to a new environment. In this work we consider a scenario in which the target environment has been scanned to obtain a 3D model of the scene suitable to generate large quantities of synthetic data automatically paired with localization labels. We hence investigate the use of Unsupervised Domain Adaptation techniques exploiting labeled synthetic data and unlabeled real data to train localization algorithms. To carry out the study, we introduce a new dataset composed of synthetic and real images labeled with their 6-DOF poses collected in four different indoor rooms. A new method based on self-supervision and attention modules is hence proposed and tested on the proposed dataset. Results show that our method improves over baselines and state-of-the-art algorithms tackling similar tasks.

Dataset

We propose a dataset of synthetic and real images related to 4 rooms of "Galleria Regionale Palazzo Bellomo" located in Siracusa, Italy. The dataset contains two set of images, synthetic and real which are divided has follows:

	Real			Simulated
	Train	Test	Val	Train	Test	Val
Room 1	561	373	252	8221	4078	4154
Room 2	562	305	233	6299	3280	3081
Room 3	405	253	321	10493	3204	3493
Room 4	128	88	65	2049	1096	989
Total	1656	1019	871	27062	11658	11718

Methods

We investigate the novel task of unsupervised domain adaptation for 6DOF visual localization inindoor scenarios.
We propose a first dataset to study the consideredproblem. The dataset has been acquired in 4 different rooms of a cultural heritage site and contains synthetic and real data which has been labeled with 6-DOF camera poses for algorithms evaluation and comparison. We publicly release the dataset to encourage research in this domain.
We propose a new approach based on self-supervision and attention modules that outperforms baselines state-of-the-art approaches

Video

Paper

Daniele Di Mauro, Antonino Furnari, Giovanni Maria Farinella, Unsupervised Domain Adaptation for 6DOF Indoor Localization, VISAPP 2021 Paper

@inproceedings{dimauro2020synthetic,
    title={Unsupervised Domain Adaptation for 6DOF Indoor Localization},
    author={Di Mauro, Daniele and Furnari, Antonino and Signorello, Giovanni and Farinella, Giovanni Maria},
    booktitle={VISAPP},
    year={2021}
}

Acknowledgement

This research is supported by XENIA Progetti s.r.l., by project VALUE - Visual Analysis for Localization and Understanding of Environments (N. 08CT6209090207 - CUP G69J18001060007) - PO FESR 2014/2020 - Azione 1.1.5., by Piano della Ricerca 2016-2018 linea di Intervento 1 CHANCE - University of Catania. The authors would like to thank Regione Siciliana Assessorato dei Beni Culturali dell'Identità Siciliana - Dipartimento dei Beni Culturali e dell'Identità Siciliana and Polo regionale di Siracusa per i siti culturali - Galleria Regionale di Palazzo Bellomo.