InLoc: Indoor Visual Localization with Dense Matching and View Synthesis

Taira, Hajime; Okutomi, Masatoshi; Sattler, Torsten; Cimpoi, Mircea; Pollefeys, Marc; Šivic, Josef; Pajdla, Tomáš; Torii, Akihiko

doi:10.1109/tpami.2019.2952114

Cited by 73 publications

(86 citation statements)

References 88 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This shows that pose regression techniques such as PoseNet currently do not scale well. This is in line with reports from other work, which reports problems when trying to train such methods in more complex scenes [63], [68], [69].…”

Section: Relevance Of the Resultssupporting

confidence: 92%

“…However, these methods do not yet achieve the same pose accuracy as structure-based methods on outdoor scenes [67]. In addition, training them on larger datasets still seems to be an open problem [68], [69]. In this paper, we show that a state-of-the-art PoseNet variant [65] performs worse than a simple image-based baseline on a medium-scale dataset.…”

Section: Related Workmentioning

confidence: 85%

“…These methods outperform structure-based approaches in terms of pose accuracy [73]. However, they currently are not able to handle larger scenes such as the ones considered in this paper [69], [73].…”

Section: Related Workmentioning

confidence: 96%

See 2 more Smart Citations

Are Large-Scale 3D Models Really Necessary for Accurate Visual Localization?

Torii

Taira

Šivic

et al. 2021

IEEE Trans. Pattern Anal. Mach. Intell.

Self Cite

View full text Add to dashboard Cite

Accurate visual localization is a key technology for autonomous navigation. 3D structure-based methods employ 3D models of the scene to estimate the full 6 degree-of-freedom (DOF) pose of a camera very accurately. However, constructing (and extending) large-scale 3D models is still a significant challenge. In contrast, 2D image retrieval-based methods only require a database of geo-tagged images, which is trivial to construct and to maintain. They are often considered inaccurate since they only approximate the positions of the cameras. Yet, the exact camera pose can theoretically be recovered when enough relevant database images are retrieved. In this paper, we demonstrate experimentally that large-scale 3D models are not strictly necessary for accurate visual localization. We create reference poses for a large and challenging urban dataset. Using these poses, we show that combining image-based methods with local reconstructions results in a higher pose accuracy compared to state-of-the-art structure-based methods, albeight at higher run-time costs. We show that some of these run-time costs can be alleviated by exploiting known database image poses. Our results suggest that we might want to reconsider the need for large-scale 3D models in favor of more local models, but also that further research is necessary to accelerate the local reconstruction process.

show abstract

Section: Relevance Of the Resultssupporting

confidence: 92%

Section: Related Workmentioning

confidence: 85%

See 1 more Smart Citation

Are Large-Scale 3D Models Really Necessary for Accurate Visual Localization?

Torii

Taira

Šivic

et al. 2021

IEEE Trans. Pattern Anal. Mach. Intell.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Indoor localization has always been an urgently needed service in our society, which can be used for indoor navigation, daily activities tracking and many other amazing applications [1], [2]. Compared with outdoor localization, indoor localization, without GPS signal, faces many challenges.…”

Section: Introductionmentioning

confidence: 99%

Online Learning-Based WIFI Radio Map Updating Considering High-Dynamic Environmental Factors

Niu¹,

Zhang

Wang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

Indoor localization has been recognized as a promising research around the world, and fingerprint-based localization method which leverages WIFI Received Signal Strength (RSS) has been extensively studied since widespread deployment of Access Points (APs) makes WIFI signals omnipresent and easily be obtained. A primary weakness of WIFI-based fingerprinting localization approach lies in its vulnerability under environmental changes and alteration of AP deployment. Despite some studies focus on dealing with effects of AP alterations and low-dynamic environmental factors, such as humidity, temperature, etc., influences of high-dynamic factors, such as changes of crowds' density and position, on WIFI radio map have not been sufficiently studied. In this work, we propose OWUH, an Online Learning-based WIFI Radio Map Updating service considering influences of high-dynamic factors. OWUH utilizes sensors in smart phones as the source of RSS datasets, and it combines historical and newly collected RSS data and purposeful probe data as dataset to incrementally update radio map, which means, compared with traditional methods, the OWUH approach requires a smaller number of RSS data for frequent updating of radio map. Moreover, in order to further enlarge our dataset, we take static data and low-dynamic data into account. An improved online learning method is proposed to recognize periodic pattern and update current radio map. Extensive experiments with 15 volunteers across 10 days indicate that OWUH effectively accommodates RSS variations over time and derives accurate prediction of fresh radio map with mean errors of less than 5dB, outperforming existing approaches. INDEX TERMS Indoor localization, WIFI radio map, crowdsourcing, online learning, high-dynamic.

show abstract

“…The state of the art on online camera localisation [2,7] usually relies on global and local features indexing combined with costly verification steps. [7] generates virtual camera view point from a dense 3D model in order to verify the retrieved pose. In our work, the refinement step relies on the alignment of point clouds, so we do not need to construct a costly 3D dense model, neither to generate artificial data.…”

Section: Introductionmentioning

confidence: 99%

Geometric Camera Pose Refinement with Learned Depth Maps

Piasco

Sidibé

Demonceaux

et al. 2019

2019 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

We present a new method for image-only camera relocalisation composed of a fast image indexing retrieval step followed by pose refinement based on ICP (Iterative Closest Point). The first step aims to find an initial pose for the query by evaluating images similarity with low dimensional global deep descriptors. Subsequently, we predict with a fully convolutional deep encoder-decoder neural network a dense depth map from the image query. We use this depth map to create a local point cloud and refine the initial query pose using an ICP algorithm. We demonstrate the effectiveness of our new approach on various indoor scenes. Compared to learned pose regression methods, our proposal can be used on multiple scenes without the need of a specific weights-setup for each scene, while showing equivalent results.

show abstract

InLoc: Indoor Visual Localization with Dense Matching and View Synthesis

Cited by 73 publications

References 88 publications

Are Large-Scale 3D Models Really Necessary for Accurate Visual Localization?

Are Large-Scale 3D Models Really Necessary for Accurate Visual Localization?

Online Learning-Based WIFI Radio Map Updating Considering High-Dynamic Environmental Factors

Geometric Camera Pose Refinement with Learned Depth Maps

Contact Info

Product

Resources

About