SemSegMap – 3D Segment-based Semantic Localization

Cramariuc, Andrei; Tschopp, Florian; Alatur, Nikhilesh; Benz, Stefan; Falck, Tillmann; Brühlmeier, Marius; Hahn, Benjamin; Nieto, Juan; Siegwart, Roland

doi:10.1109/iros51168.2021.9636156

Cited by 19 publications

(4 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Building on this research, Ratz et al [18] showed that lidar segments fused with visual data further improve the performance of global localization algorithms. Cramariuc et al [5] fused both colour and semantic information from images to create an enriched point cloud that was later segmented and used for localization. We use this segment-direct concept as the basis of InstaLoc, however in contrast to these approaches, InstaLoc does not use engineered segmentation methods, nor images, to extract the semantic information but directly learns to predict the perpoint instance annotation.…”

Section: A Outdoor Segment-based Localizationmentioning

confidence: 99%

InstaLoc: One-shot Global Lidar Localisation in Indoor Environments through Instance Learning

Zhang,

Digumarti,

Tinchev

et al. 2023

Robotics: Science and Systems XIX

View full text Add to dashboard Cite

Localization for autonomous robots in prior maps is crucial for their functionality. This paper offers a solution to this problem for indoor environments called InstaLoc, which operates on an individual lidar scan to localize it within a prior map. We draw on inspiration from how humans navigate and position themselves by recognizing the layout of distinctive objects and structures. Mimicking the human approach, InstaLoc identifies and matches object instances in the scene with those from a prior map. As far as we know, this is the first method to use panoptic segmentation directly inferring on 3D lidar scans for indoor localization. InstaLoc operates through two networks based on spatially sparse tensors to directly infer dense 3D lidar point clouds. The first network is a panoptic segmentation network that produces object instances and their semantic classes. The second smaller network produces a descriptor for each object instance. A consensus based matching algorithm then matches the instances to the prior map and estimates a six degrees of freedom (DoF) pose for the input cloud in the prior map. InstaLoc utilizes two efficient networks, requires only one to two hours of training on a mobile GPU, and runs in real-time at 1 Hz. Our method achieves between two and four times more detections when localizing, as compared to baseline methods, and achieves higher precision on these detections.

show abstract

Section: A Outdoor Segment-based Localizationmentioning

confidence: 99%

InstaLoc: One-shot Global Lidar Localisation in Indoor Environments through Instance Learning

Zhang,

Digumarti,

Tinchev

et al. 2023

Robotics: Science and Systems XIX

View full text Add to dashboard Cite

show abstract

“…s3li mapping instead tests the ability of SLAM algorithms to produce a complete and consistent map of the environment relying on stereo as well as LiDAR depth measurements, as a confined area is explored multiple times and there should be plenty of loop closure opportunities. Finally, s3li landmarks allows testing of segmentation techniques for natural landmarks, with the aim of either semantic mapping or segment-based loop closure detection [31]. An overview of the sequences, including D-GNSS tracks, is provided in Table II.…”

Section: A Sequencesmentioning

confidence: 99%

Challenges of SLAM in Extremely Unstructured Environments: The DLR Planetary Stereo, Solid-State LiDAR, Inertial Dataset

Giubilato

Stürzl

Wedler

et al. 2022

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

We present the DLR Planetary Stereo, Solid-State LiDAR, Inertial (S3LI) dataset, recorded on Mt. Etna, Sicily, an environment analogous to the Moon and Mars, using a handheld sensor suite with attributes suitable for implementation on a space-like mobile rover. The environment is characterized by challenging conditions regarding both the visual and structural appearance: severe visual aliasing poses significant limitations to the ability of visual SLAM systems to perform place recognition, while the absence of outstanding structural details, joined with the limited Field-of-View of the utilized Solid-State LiDAR sensor, challenges traditional LiDAR SLAM for the task of pose estimation using point clouds alone. With this data, that covers more than 4 kilometers of travel on soft volcanic slopes, we aim to: 1) provide a tool to expose limitations of state-ofthe-art SLAM systems with respect to environments, which are not present in widely available datasets and 2) motivate the development of novel localization and mapping approaches, that rely efficiently on the complementary capabilities of the two sensors. The dataset is accessible at the following url: https://rmc.dlr.de/s3li_dataset

show abstract

“…Li et al [17] use semantics to enhance SC and proposed semantic scan context. Similarly, Cramariuc et al [9] utilize semantic information to improve the performance of SegMatch. In contrast to these semanticenhanced methods, our approach only uses depth information of the range images to achieve online performance and is easier to generalize to different environments and datasets collected by different LiDAR sensors.…”

Section: Related Workmentioning

confidence: 99%

OverlapTransformer: An Efficient and Rotation-Invariant Transformer Network for LiDAR-Based Place Recognition

Ma¹,

Zhang²,

Xu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Place recognition is an important capability for autonomously navigating vehicles operating in complex environments and under changing conditions. It is a key component for tasks such as loop closing in SLAM or global localization. In this paper, we address the problem of place recognition based on 3D LiDAR scans recorded by an autonomous vehicle. We propose a novel lightweight neural network exploiting the range image representation of LiDAR sensors to achieve fast execution with less than 4 ms per frame. Based on that, we design a yaw-rotation-invariant architecture exploiting a transformer network, which boosts the place recognition performance of our method. We evaluate our approach on the KITTI and Ford Campus datasets. The experimental results show that our method can effectively detect loop closures compared to the state-of-the-art methods and generalizes well across different environments. To further evaluate long-term place recognition performance, we provide a novel challenging Haomo dataset, which contains LiDAR sequences recorded by a mobile robot in repetitive places across seasons. Both the implementation of our method and our new Haomo dataset are released here: https://github.com/haomo-ai/OverlapTransformer

show abstract

SemSegMap – 3D Segment-based Semantic Localization

Cited by 19 publications

References 37 publications

InstaLoc: One-shot Global Lidar Localisation in Indoor Environments through Instance Learning

InstaLoc: One-shot Global Lidar Localisation in Indoor Environments through Instance Learning

Challenges of SLAM in Extremely Unstructured Environments: The DLR Planetary Stereo, Solid-State LiDAR, Inertial Dataset

OverlapTransformer: An Efficient and Rotation-Invariant Transformer Network for LiDAR-Based Place Recognition

Contact Info

Product

Resources

About