3D object detection in voxelized point cloud scene

Table 1. Accuracy comparison of max pooling layer and average pooling layer in 3D object detection of car AP40
View table
View in Article
Table 1. Accuracy comparison of max pooling layer and average pooling layer in 3D object detection of car AP40
Easy Mod Hard
Max 92.34 84.63 82.48
Average 92.17 83.01 82.27

Table 2. Comparison of car AP40 effects of different neighborhood thresholds on multi-scale ROI feature aggregation module in 3D object detection
View table
View in Article
Table 2. Comparison of car AP40 effects of different neighborhood thresholds on multi-scale ROI feature aggregation module in 3D object detection
［block3］［block4］ Easy Mod Hard
［1-2］［2，4］ 92.42 85.43 83.04
［2，4］［2，4］ 92.34 84.63 82.48
［2，4］［4，8］ 92.37 85.15 82.98

Table 3. Performance comparison on the KITTI test set with AP calculated by 11 recall positions of 3D object

View table

View in Article

Table 3. Performance comparison on the KITTI test set with AP calculated by 11 recall positions of 3D object

	Modality	FPS	Car	Pedestrian	Cyclist
Easy	Modality	Mod	Hard	Easy	Mod	Hard	Easy	Mod	Hard
MV3D	Lidar & Img.	3	71.19	56.6	55.3	N/A	N/A	N/A	N/A	N/A	N/A
SECOND	Lidar	30.4	87.43	76.48	69.1	N/A	N/A	N/A	N/A	N/A	N/A
VoxelNet	Lidar	4.44	81.97	65.46	62.85	57.86	53.42	48.87	67.17	47.65	45.11
Voxel R-CNN	Lidar	25.2	89.41	84.52	78.93	N/A	N/A	N/A	N/A	N/A	N/A
PointPillars	Lidar	62.5	85.22	76.52	69.73	63.87	58.08	52.8	77.16	59.22	55.91
Pillar RCNN	Lidar	40.3	89.26	83.61	78.58	65.03	58.80	54.47	85.35	72.56	68.68

Table 4. Performance comparison on the KITTI test set with AP calculated by 11 recall positions of BEV object

View table

View in Article

Table 4. Performance comparison on the KITTI test set with AP calculated by 11 recall positions of BEV object

	Modality	FPS	Car	Pedestrian	Cyclist
Easy	Modality	Mod	Hard	Easy	Mod	Hard	Easy	Mod	Hard
MV3D	Lidar & Img.	3	86.18	77.32	76.33	N/A	N/A	N/A	N/A	N/A	N/A
PIXOR	Lidar & Img.	10.6	86.79	80.75	76.6	N/A	N/A	N/A	N/A	N/A	N/A
SECOND	Lidar	30.4	89.96	87.07	79.66	N/A	N/A	N/A	N/A	N/A	N/A
VoxelNet	Lidar	4.44	89.6	84.81	78.57	69.95	61.05	56.98	74.41	52.18	50.49
PointPillars	Lidar	62.5	90.07	87.44	84.78	71.18	65.92	61.36	82.2	62.81	59.77
Pillar RCNN	Lidar	40.3	90.15	88.13	87.65	66.13	61.16	58.03	92.21	73.85	70.65

Tools

Get Citation

Copy Citation Text

Rui-long LI, Chuan WU, Ming ZHU. 3D object detection in voxelized point cloud scene[J]. Chinese Journal of Liquid Crystals and Displays, 2022, 37(10): 1355

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Research Articles

Received: Mar. 10, 2022

Accepted: --

Published Online: Oct. 10, 2022

The Author Email: Chuan WU (wuchuan0458@sina.com)

DOI:10.37188/CJLCD.2022-0082

Topics

Table 1. Accuracy comparison of max pooling layer and average pooling layer in 3D object detection of car AP40

Table 1. Accuracy comparison of max pooling layer and average pooling layer in 3D object detection of car AP40

Table 2. Comparison of car AP40 effects of different neighborhood thresholds on multi-scale ROI feature aggregation module in 3D object detection

Table 2. Comparison of car AP40 effects of different neighborhood thresholds on multi-scale ROI feature aggregation module in 3D object detection

Table 3. Performance comparison on the KITTI test set with AP calculated by 11 recall positions of 3D object

Table 3. Performance comparison on the KITTI test set with AP calculated by 11 recall positions of 3D object

Table 4. Performance comparison on the KITTI test set with AP calculated by 11 recall positions of BEV object

Table 4. Performance comparison on the KITTI test set with AP calculated by 11 recall positions of BEV object