MASS: Multi-Attentional Semantic Segmentation of LiDAR Data for Dense Top-View Understanding

Peng, Kunyu; Fei, Juncong; Yang, Kailun; Roitberg, Alina; Zhang, Jiaming; Bieder, Frank; Heidenreich, Philipp; Stiller, Christoph; Stiefelhagen, Rainer

doi:10.1109/tits.2022.3145588

Cited by 36 publications

(8 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, Google’s Waymo uses LiDAR for detecting unexpected objects [ 27 ]. To accurately classify and recognize objects from data detected by LiDAR, various models have been developed, and algorithm-development studies using semantic segmentation have been conducted [ 28 , 29 ]. Moreover, studies have been investigating the detectability of objects by LiDAR.…”

Section: Literature Reviewmentioning

confidence: 99%

Empirical Analysis of Autonomous Vehicle’s LiDAR Detection Performance Degradation for Actual Road Driving in Rain and Fog

Kim

Park

2023

Sensors

View full text Add to dashboard Cite

Light detection and ranging (LiDAR) is widely used in autonomous vehicles to obtain precise 3D information about surrounding road environments. However, under bad weather conditions, such as rain, snow, and fog, LiDAR-detection performance is reduced. This effect has hardly been verified in actual road environments. In this study, tests were conducted with different precipitation levels (10, 20, 30, and 40 mm/h) and fog visibilities (50, 100, and 150 m) on actual roads. Square test objects (60 × 60 cm2) made of retroreflective film, aluminum, steel, black sheet, and plastic, commonly used in Korean road traffic signs, were investigated. Number of point clouds (NPC) and intensity (reflection value of points) were selected as LiDAR performance indicators. These indicators decreased with deteriorating weather in order of light rain (10–20 mm/h), weak fog (<150 m), intense rain (30–40 mm/h), and thick fog (≤50 m). Retroreflective film preserved at least 74% of the NPC under clear conditions with intense rain (30–40 mm/h) and thick fog (<50 m). Aluminum and steel showed non-observation for distances of 20–30 m under these conditions. ANOVA and post hoc tests suggested that these performance reductions were statistically significant. Such empirical tests should clarify the LiDAR performance degradation.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Empirical Analysis of Autonomous Vehicle’s LiDAR Detection Performance Degradation for Actual Road Driving in Rain and Fog

Kim

Park

2023

Sensors

View full text Add to dashboard Cite

show abstract

“…Although these learning-based instance segmentation methods can accurately extract objects, they cannot recognize the ground or walls that are also important for autonomous navigation. Unlike the instance segmentation, others proposed the semantic segmentation method that can label the point-wise class [18]- [20]. These methods extract the ground without considering traversability.…”

Section: Learning-based Segmentationmentioning

confidence: 99%

TRAVEL: Traversable Ground and Above-Ground Object Segmentation Using Graph Representation of 3D LiDAR Scans

Jung

Lim

et al. 2022

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

Perception of traversable regions and objects of interest from a 3D point cloud is one of the critical tasks in autonomous navigation. A ground vehicle needs to look for traversable terrains that are explorable by wheels. Then, to make safe navigation decisions, the segmentation of objects positioned on those terrains has to be followed up. However, over-segmentation and under-segmentation can negatively influence such navigation decisions. To that end, we propose TRAVEL, which performs traversable ground detection and object clustering simultaneously using the graph representation of a 3D point cloud. To segment the traversable ground, a point cloud is encoded into a graph structure, tri-grid field, which treats each tri-grid as a node. Then, the traversable regions are searched and redefined by examining local convexity and concavity of edges that connect nodes. On the other hand, our above-ground object segmentation employs a graph structure by representing a group of horizontally neighboring 3D points in a spherical-projection space as a node and vertical/horizontal relationship between nodes as an edge. Fully leveraging the node-edge structure, the above-ground segmentation ensures real-time operation and mitigates over-segmentation. Through experiments using simulations, urban scenes, and our own datasets, we have demonstrated that our proposed traversable ground segmentation algorithm outperforms other state-ofthe-art methods in terms of the conventional metrics and that our newly proposed evaluation metrics are meaningful for assessing the above-ground segmentation. We will make the code and our own dataset available to public at https://github.com/url-kaist/TRAVEL.

show abstract

“…lem is to use multi-camera or LiDAR systems [8], [9], [10]. These systems, however, introduce a further level of complexity as they either need to stitch images of multiple camera sources or resort to the combination of the RGB image information with LiDAR views in order to get a holistic view of the entire surrounding [11].…”

Section: Robust Model Baseline Modelmentioning

confidence: 99%

Panoramic Panoptic Segmentation: Towards Complete Surrounding Understanding via Unsupervised Contrastive Learning

Jaus

Yang

Stiefelhagen

2021

2021 IEEE Intelligent Vehicles Symposium (IV)

Self Cite

View full text Add to dashboard Cite

In this work, we introduce panoramic panoptic segmentation as the most holistic scene understanding both in terms of field of view and image level understanding. A complete surrounding understanding provides a maximum of information to the agent, which is essential for any intelligent vehicle in order to make informed decisions in a safetycritical dynamic environment such as real-world traffic. In order to overcome the lack of annotated panoramic images, we propose a framework which allows model training on standard pinhole images and transfers the learned features to a different domain. Using our proposed method, we manage to achieve significant improvements of over 5% measured in PQ over nonadapted models on our Wild Panoramic Panoptic Segmentation (WildPPS) dataset. We show that our proposed Panoramic Robust Feature (PRF) framework is not only suitable to improve performance on panoramic images but can be beneficial whenever model training and deployment are executed on data taken from different distributions. As an additional contribution, we publish WildPPS: The first panoramic panoptic image dataset to foster progress in surrounding perception.

show abstract

MASS: Multi-Attentional Semantic Segmentation of LiDAR Data for Dense Top-View Understanding

Cited by 36 publications

References 59 publications

Empirical Analysis of Autonomous Vehicle’s LiDAR Detection Performance Degradation for Actual Road Driving in Rain and Fog

Empirical Analysis of Autonomous Vehicle’s LiDAR Detection Performance Degradation for Actual Road Driving in Rain and Fog

TRAVEL: Traversable Ground and Above-Ground Object Segmentation Using Graph Representation of 3D LiDAR Scans

Panoramic Panoptic Segmentation: Towards Complete Surrounding Understanding via Unsupervised Contrastive Learning

Contact Info

Product

Resources

About