SARPNET: Shape attention regional proposal network for liDAR-based 3D object detection

Ye, Yangyang; Chen, Houjin; Zhang, Chi; Hao, Xiaoli; Zhang, Zhaoxiang

doi:10.1016/j.neucom.2019.09.086

Cited by 77 publications

(24 citation statements)

References 39 publications

(86 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We calculate the distance weight ω j between p lay i and p lay j through the K-nearest neighbors. As shown in Equation (12), the further away from p lay i , the smaller the contribution on p lay i . Here, K = 3.…”

Section: Multi-scale Feature Extraction Module Based On Adaptive Feat...mentioning

confidence: 99%

“…refs. [11][12][13][14][15][16]. After voxelization, the 3D voxel or 2D bird's eye view grid can be processed by convolutional neural networks.…”

Section: Introductionmentioning

confidence: 99%

“…2022, 14, x FOR PEER REVIEW 2 of 21 not express the relevancy of the spatial-variant well. Voxel-based methods generally convert the unstructured point clouds into a 3D voxel or a 2D bird's eye view grid, such as in refs [11][12][13][14][15][16]. After voxelization, the 3D voxel or 2D bird's eye view grid can be processed by convolutional neural networks.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

AFE-RCNN: Adaptive Feature Enhancement RCNN for 3D Object Detection

Shuang

Huang

et al. 2022

Remote Sensing

View full text Add to dashboard Cite

The point clouds scanned by lidar are generally sparse, which can result in fewer sampling points of objects. To perform precise and effective 3D object detection, it is necessary to improve the feature representation ability to extract more feature information of the object points. Therefore, we propose an adaptive feature enhanced 3D object detection network based on point clouds (AFE-RCNN). AFE-RCNN is a point-voxel integrated network. We first voxelize the raw point clouds and obtain the voxel features through the 3D voxel convolutional neural network. Then, the 3D feature vectors are projected to the 2D bird’s eye view (BEV), and the relationship between the features in both spatial dimension and channel dimension is learned by the proposed residual of dual attention proposal generation module. The high-quality 3D box proposals are generated based on the BEV features and anchor-based approach. Next, we sample key points from raw point clouds to summarize the information of the voxel features, and obtain the key point features by the multi-scale feature extraction module based on adaptive feature adjustment. The neighboring contextual information is integrated into each key point through this module, and the robustness of feature processing is also guaranteed. Lastly, we aggregate the features of the BEV, voxels, and point clouds as the key point features that are used for proposal refinement. In addition, to ensure the correlation among the vertices of the bounding box, we propose a refinement loss function module with vertex associativity. Our AFE-RCNN exhibits comparable performance on the KITTI dataset and Waymo open dataset to state-of-the-art methods. On the KITTI 3D detection benchmark, for the moderate difficulty level of the car and the cyclist classes, the 3D detection mean average precisions of AFE-RCNN can reach 81.53% and 67.50%, respectively.

show abstract

Section: Multi-scale Feature Extraction Module Based On Adaptive Feat...mentioning

confidence: 99%

“…refs. [11][12][13][14][15][16]. After voxelization, the 3D voxel or 2D bird's eye view grid can be processed by convolutional neural networks.…”

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

AFE-RCNN: Adaptive Feature Enhancement RCNN for 3D Object Detection

Shuang

Huang

et al. 2022

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Non-local [13] modeling of long-range dependencies is proven effective in 2D object detection. In 3D object detection, [14] forms attention fusion in both bird's eye view and vertical view. [15] designs a triple attention fusion of channel-wise, point-wise, and voxel-wise features.…”

Section: Introductionmentioning

confidence: 99%

PLNL-3DSSD: Part-Aware 3D Single Stage Detector Using Local And Non-Local Attention

Liu

Chen

et al. 2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

3D object detection in the real crowded scene is still a challenging task due to occlusion and density change. We propose a part-aware 3D single-stage detector with local and non-local attention (PLNL-3DSSD) to fully use part information and inter-object relation. A primary part feature fusion is proposed for encoding the entire box feature vector by introducing semantic parts dividing. We develop a parallel part branch for robust and accurate object detection. We also develop local and non-local attention in set abstraction for enhancing data flow transfer between objects. Our method ranks second in single-stage 3D object detector on the KITTI 3D car detection benchmark while ensuring satisfactory efficiency.

show abstract

“…In this work only pedestrian areas are analyzed, so it is not necessary to segment urban ground elements, because citizens can navigate freely through squares and areas closed to traffic, although people in wheelchairs are limited in their mobility because of steps. Nor is it necessary to differentiate the classes of objects, a topic extensively studied [17][18][19], since this works focuses on the shadow they produce. This paper is structured as follows.…”

Section: Introductionmentioning

confidence: 99%

Generation and Enrichment of Pedestrian Maps with Vertical Shadows in Urban Environments from Mobile Laser Scanning Data

Balado

Díaz-Vilariño

Frías

et al. 2020

JASTT

View full text Add to dashboard Cite

Cities are becoming more pedestrian-friendly, reducing traffic and promoting physical activity and walking. However, prolonged exposure to the sun can cause sunburn and skin problems, so minimizing exposure to the sun while travelling is especially relevant at certain latitudes and in the summer months. This paper proposes a method for modelling urban contours and generating pedestrian maps with the location of shaded areas and accessibility barriers. The proposed method uses as input data a point cloud of an urban environment acquired with Mobile Laser Scanning. First, the input point cloud is segmented in ground points, obstacle points, and points causing shadows. Then, the three segmented point clouds are rasterized and the corresponded images are combined to obtain the navigable ground and the shaded areas. Finally, from the navigable ground, a navigation map is generated for pedestrians. To check the usefulness of this navigation map, a pathfinding algorithm is applied. The results show a correct generation of the navigable ground, and routes prioritizing the trajectory by shadow areas. Depending on the weighting between sun and shaded areas, the routes obtained show differences in distance travelled and sun exposure. The proposed method is sensitive to the existence of obstacles and noise in the point clouds.

show abstract

SARPNET: Shape attention regional proposal network for liDAR-based 3D object detection

Cited by 77 publications

References 39 publications

AFE-RCNN: Adaptive Feature Enhancement RCNN for 3D Object Detection

AFE-RCNN: Adaptive Feature Enhancement RCNN for 3D Object Detection

PLNL-3DSSD: Part-Aware 3D Single Stage Detector Using Local And Non-Local Attention

Generation and Enrichment of Pedestrian Maps with Vertical Shadows in Urban Environments from Mobile Laser Scanning Data

Contact Info

Product

Resources

About