Data Association Aware Semantic Mapping and Localization via a Viewpoint-Dependent Classifier Model

Turetsky, Vladimir; Feldman, Yuri; Indelman, Vadim

doi:10.1109/iros40897.2019.8967855

Cited by 15 publications

(22 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…is the prior over object poses newly observed at time k. As opposed to [7], this formulation also supports an increasing number of objects known at each time step, with both X o,r k and C r k increasing in dimension. Note that in general b r k is different for each class realization, as models (1) are different for each class.…”

Section: Local Hybrid Belief Maintenancementioning

confidence: 94%

“…Feldman and Indelman [6] proposed a sequential object classification that utilizes a viewpoint dependent classifier with known relative poses a-priori. Tchuiev et al [7] maintained a hybrid belief with a viewpoint dependent classifier to disambiguate between data association realizations. These works, [7], address only sequential classification and do not consider the coupled problem with SLAM.…”

Section: Related Workmentioning

confidence: 99%

“…Additionally, we use a viewpoint-dependent classifier model that "predicts" classification scores (a vector of class probabilities). This model couples classifier scores with viewpoint dependency between object and camera; this coupling can be used to improve pose inference performance [7]. The viewpoint dependency is modeled as a Gaussian with parameters that depend on the relative viewpoint from the camera to the object x o,r x r k and object's class c:…”

Section: Notations and Problem Formulationmentioning

confidence: 99%

“…Our formulation for maintaining local hybrid beliefs builds upon our previous work [7], with the main differences being that here we assume the DA is solved, and the number of objects is unknown a-priori. In this section we present an overview of this approach.…”

Section: Local Hybrid Belief Maintenancementioning

confidence: 99%

See 3 more Smart Citations

Distributed Consistent Multi-Robot Semantic Localization and Mapping

Turetsky

Indelman

2020

IEEE Robot. Autom. Lett.

Self Cite

View full text Add to dashboard Cite

We present an approach for multi-robot consistent distributed localization and semantic mapping in an unknown environment, considering scenarios with classification ambiguity, where objects' visual appearance generally varies with viewpoint. Our approach addresses such a setting by maintaining a distributed posterior hybrid belief over continuous localization and discrete classification variables. In particular, we utilize a viewpoint-dependent classifier model to leverage the coupling between semantics and geometry. Moreover, our approach yields a consistent estimation of both continuous and discrete variables, with the latter being addressed for the first time, to the best of our knowledge. We evaluate the performance of our approach in a multi-robot semantic SLAM simulation and in a real-world experiment, demonstrating an increase in both classification and localization accuracy compared to maintaining a hybrid belief using local information only.

show abstract

Section: Local Hybrid Belief Maintenancementioning

confidence: 94%

Section: Related Workmentioning

confidence: 99%

Section: Notations and Problem Formulationmentioning

confidence: 99%

Section: Local Hybrid Belief Maintenancementioning

confidence: 99%

See 2 more Smart Citations

Distributed Consistent Multi-Robot Semantic Localization and Mapping

Turetsky

Indelman

2020

IEEE Robot. Autom. Lett.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Graphs are a natural representation for these types of representations. The nodes of the graph are usually given by feature or landmarks detected in the observation, from which a kernel can be applied to extract a descriptor to be matched against other observations [10], [11]. GLARE [12], its rotation invariant extension GLAROT [13], and its 3D extension [14] first compute a descriptor for each observation based on the relative distances and angles between landmarks.…”

Section: Introductionmentioning

confidence: 99%

Place Recognition in Forests With Urquhart Tessellations

Nardari

Cohen

Chen

et al. 2021

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

In this letter we present a novel descriptor based on polygons derived from Urquhart tessellations on the position of trees in a forest detected from lidar scans. We present a framework that leverages these polygons to generate a signature that is used detect previously seen observations even with partial overlap and different levels of noise while also inferring landmark correspondences to compute an affine transformation between observations. We run loop-closure experiments in simulation and real-world data map-merging from different flights of an Unmanned Aerial Vehicle (UAV) in a pine tree forest and show that our method outperforms state-of-the-art approaches in accuracy and robustness.

show abstract

Spatially-dependent Bayesian semantic perception under model and localization uncertainty

Feldman

Indelman

2020

Auton Robot

View full text Add to dashboard Cite

Data Association Aware Semantic Mapping and Localization via a Viewpoint-Dependent Classifier Model

Cited by 15 publications

References 13 publications

Distributed Consistent Multi-Robot Semantic Localization and Mapping

Distributed Consistent Multi-Robot Semantic Localization and Mapping

Place Recognition in Forests With Urquhart Tessellations

Spatially-dependent Bayesian semantic perception under model and localization uncertainty

Contact Info

Product

Resources

About