Exploiting Multi-Layer Grid Maps for Surround-View Semantic Segmentation of Sparse LiDAR Data

Bieder, Frank; Wirges, Sascha; Janosovits, Johannes; Richter, Sven; Wang, Zheyuan; Stiller, Christoph

doi:10.1109/iv47402.2020.9304848

Cited by 20 publications

(45 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is expected as the used stereo disparity estimation is more accurate than the monocular depth estimation. In general, the numbers for both setups are in similar regions as the ones presented in the Lidar-based semantic grid map estimation from Bieder et al in [30]. They reach a 39.8% mean IoU with their best configuration.…”

Section: Intersection Over Unionsupporting

confidence: 68%

“…In our quantitative evaluation, we showed the benefits of our evidential model by obtaining significantly better error metrics when considering the uncertainties. This is one of the main advantages of our method compared to other publications and enables our pipeline to perform comparably well to competitive ones using more expensive sensors such as Lidar [14,30]. The second advantage is the underlying semantic evidential representation that makes fusion with other sensor types as range sensors straight forward, see [1].…”

Section: Discussionmentioning

confidence: 83%

“…Experiments showed that CR can be improved by up to 10% depending on the sequence when merging the two classes. Besides the Lidar-based semantic top-view maps presented in [30], we can compare our results to the hybrid approach using Lidar and RGB images from Erkent et al presented in [14]. They achieve a ratio of correctly labeled cells of 81% in their best performing setup, indicating that our approach performs slightly better.…”

Section: Intersection Over Unionmentioning

confidence: 82%

See 2 more Smart Citations

Semantic Evidential Grid Mapping Using Monocular and Stereo Cameras

Richter

Wang

Beck³

et al. 2021

Sensors

Self Cite

View full text Add to dashboard Cite

Accurately estimating the current state of local traffic scenes is one of the key problems in the development of software components for automated vehicles. In addition to details on free space and drivability, static and dynamic traffic participants and information on the semantics may also be included in the desired representation. Multi-layer grid maps allow the inclusion of all of this information in a common representation. However, most existing grid mapping approaches only process range sensor measurements such as Lidar and Radar and solely model occupancy without semantic states. In order to add sensor redundancy and diversity, it is desired to add vision-based sensor setups in a common grid map representation. In this work, we present a semantic evidential grid mapping pipeline, including estimates for eight semantic classes, that is designed for straightforward fusion with range sensor data. Unlike other publications, our representation explicitly models uncertainties in the evidential model. We present results of our grid mapping pipeline based on a monocular vision setup and a stereo vision setup. Our mapping results are accurate and dense mapping due to the incorporation of a disparity- or depth-based ground surface estimation in the inverse perspective mapping. We conclude this paper by providing a detailed quantitative evaluation based on real traffic scenarios in the KITTI odometry benchmark dataset and demonstrating the advantages compared to other semantic grid mapping approaches.

show abstract

Section: Intersection Over Unionsupporting

confidence: 68%

Section: Discussionmentioning

confidence: 83%

Section: Intersection Over Unionmentioning

confidence: 82%

See 1 more Smart Citation

Semantic Evidential Grid Mapping Using Monocular and Stereo Cameras

Richter

Wang

Beck³

et al. 2021

Sensors

Self Cite

View full text Add to dashboard Cite

show abstract

“…Since the hand-crafted grid map feature extraction in [5] can result in a potential information loss, we propose to use PointNet [6] to learn features directly from the point cloud and avoid this potential information loss. In this paper, we propose a novel end-to-end method named PillarSegNet to approach dense semantic grid map estimation using sparse LiDAR data.…”

Section: Predictionmentioning

confidence: 99%

“…While most existing approaches [2], [3], [4] predict pointwise semantic scores from the sparse LiDAR point cloud, Bieder et al [5] transform the sparse LiDAR point cloud into a multi-layer grid map representation to obtain a dense topview segmentation of the LiDAR measurements. In Fig.…”

Section: Introductionmentioning

confidence: 99%