Compressed Voxel-Based Mapping Using Unsupervised Learning

Canelhas, Daniel Ricao; Schaffernicht, Erik; Stoyanov, Todor; Lilienthal, Achim J.; Davison, Andrew J.

doi:10.3390/robotics6030015

Cited by 9 publications

(8 citation statements)

References 22 publications

(24 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Brock et al (2016) also present successful results using variational autoencoders for reconstructing voxelized 3D data. Different configurations of encoding and decoding networks have also been proposed for achieving localization and for reconstructing and completing 3D shapes and environments (Dai et al, 2017; Elbaz et al, 2017; Guizilini and Ramos, 2017; Ricao Canelhas et al, 2017; Schönberger et al, 2018; Varley et al, 2017).…”

Section: Related Workmentioning

confidence: 99%

SegMap: Segment-based mapping and localization using data-driven descriptors

Dubé

Cramariuc

Dugas

et al. 2019

The International Journal of Robotics Research

166

View full text Add to dashboard Cite

Precisely estimating a robot’s pose in a prior, global map is a fundamental capability for mobile robotics, e.g., autonomous driving or exploration in disaster zones. This task, however, remains challenging in unstructured, dynamic environments, where local features are not discriminative enough and global scene descriptors only provide coarse information. We therefore present SegMap: a map representation solution for localization and mapping based on the extraction of segments in 3D point clouds. Working at the level of segments offers increased invariance to view-point and local structural changes, and facilitates real-time processing of large-scale 3D data. SegMap exploits a single compact data-driven descriptor for performing multiple tasks: global localization, 3D dense map reconstruction, and semantic information extraction. The performance of SegMap is evaluated in multiple urban driving and search and rescue experiments. We show that the learned SegMap descriptor has superior segment retrieval capabilities, compared with state-of-the-art handcrafted descriptors. As a consequence, we achieve a higher localization accuracy and a 6% increase in recall over state-of-the-art handcrafted descriptors. These segment-based localizations allow us to reduce the open-loop odometry drift by up to 50%. SegMap is open-source available along with easy to run demonstrations.

show abstract

Section: Related Workmentioning

confidence: 99%

SegMap: Segment-based mapping and localization using data-driven descriptors

Dubé

Cramariuc

Dugas

et al. 2019

The International Journal of Robotics Research

166

View full text Add to dashboard Cite

show abstract

“…While SVOs only encode binary occupancy of voxels, the Truncated Signed Distance Field (TSDF) representation (Curless and Levoy, 1996) can be used to implicitly encode surfaces in a set of voxels. Recent work has shown that bricks of TSDF voxels can be compressed effectively by converting them into a lower dimensional latent space, using Principal Component Analysis (PCA) (Canelhas et al, 2017, Tang et al, 2018 or by training an encoder-decoder neural network (Tang et al, 2020).…”

Section: Efficient Visualization Of 4d Modelsmentioning

confidence: 99%

TOWARDS FREE-VIEWPOINT VIDEO CAPTURE IN CHALLENGING ENVIRONMENTS FOR COLLABORATIVE & IMMERSIVE ANALYSIS

Frolov

Rendle

Kreskowski

et al. 2021

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. The ability to capture and explore complex real-world dynamic scenes is crucial for their detailed analysis. Tools which allow retrospective exploration of such scenes may support training of new employees or be used to evaluate industrial processes. In our work, we share insights and practical details for end-to-end acquisition of Free-Viewpoint Videos (FVV) in challenging environments and their potential for exploration in collaborative immersive virtual environments. Our lightweight capturing approach makes use of commodity DSLR cameras and focuses on improving both density and accuracy of Structure-from-Motion (SfM) reconstructions from small sets of images under difficult conditions. The integration of captured 3D models over time into a compact representation allows for efficient visualization of detailed FVVs in an immersive multi-user virtual reality system. We demonstrate our workflow on a representative acquisition of a suction excavation process and outline a use-case for exploration and interaction between collocated users and the FVV in a collaborative virtual environment.

show abstract

“…Instead, we build a local PCA-SDF shape representation, which represents each local shape as a linear combination of implicit volumetric prototypes, leading to finer details and order of magnitude faster inference than prior works. Similar eigenanalysis of SDF has been applied for either global shape representation [67] or geometry compression [68,69]. However, none of them develop their algorithms from our perspective of local shape priors, which is motivated from the assumption that local shapes at the voxel level share similarities.…”

Section: Related Workmentioning

confidence: 99%

Voxel-based 3D Detection and Reconstruction of Multiple Objects from a Single Image

Liu¹,

Liu²

2021

Preprint

View full text Add to dashboard Cite

Inferring 3D locations and shapes of multiple objects from a single 2D image is a long-standing objective of computer vision. Most of the existing works either predict one of these 3D properties or focus on solving both for a single object. One fundamental challenge lies in how to learn an effective representation of the image that is well-suited for 3D detection and reconstruction. In this work, we propose to learn a regular grid of 3D voxel features from the input image which is aligned with 3D scene space via a 3D feature lifting operator. Based on the 3D voxel features, our novel CenterNet-3D detection head formulates the 3D detection as keypoint detection in the 3D space. Moreover, we devise an efficient coarse-to-fine reconstruction module, including coarse-level voxelization and a novel local PCA-SDF shape representation, which enables fine detail reconstruction and one order of magnitude faster inference than prior methods. With complementary supervision from both 3D detection and reconstruction, one enables the 3D voxel features to be geometry and context preserving, benefiting both tasks. The effectiveness of our approach is demonstrated through 3D detection and reconstruction in single object and multiple object scenarios. Code is available at http://cvlab.cse. msu.edu/project-mdr.html.

show abstract

Compressed Voxel-Based Mapping Using Unsupervised Learning

Cited by 9 publications

References 22 publications

SegMap: Segment-based mapping and localization using data-driven descriptors

SegMap: Segment-based mapping and localization using data-driven descriptors

TOWARDS FREE-VIEWPOINT VIDEO CAPTURE IN CHALLENGING ENVIRONMENTS FOR COLLABORATIVE & IMMERSIVE ANALYSIS

Voxel-based 3D Detection and Reconstruction of Multiple Objects from a Single Image

Contact Info

Product

Resources

About