Mask-Based and Multi-Modal Bird's Eye View Feature-Guided Point Cloud Panoptic Segmentation

Fig. 6. Comparative visualization of point cloud panoptic segmentation of different models. (a) BEVGuide-PS(L+C); (b) BEVGuide-PS(L); (c) MaskPLS; (d) EfficientLPS

Download full size

Fig. 7. Comparative visualization of ‘Thing’ attention mask. (a) BEVGuide-PS(L+C); (b) BEVGuide-PS(L); (c) MaskPLS

Download full size

Fig. 8. Comparative visualization of ‘Stuff’ attention mask. (a) BEVGuide-PS(L+C); (b) BEVGuide-PS(L); (c) MaskPLS

Download full size

Fig. 9. Camera projection view of point cloud panoptic segmentation

Download full size

Fig. 10. Detailed view of BEVGuide-PS panoptic segmentation

Download full size

Table 1. Comparison of point cloud panoptic segmentation accuracy on nuScenes dataset

View table

Table 1. Comparison of point cloud panoptic segmentation accuracy on nuScenes dataset

Model	PQ	PQ^†	RQ	SQ	PQ_Th	RQ_Th	SQ_Th	PQ_St	RQ_St	SQ_St	mIoU
EfficientLPS	62.0	65.6	73.9	83.4	56.8	68.0	83.2	70.6	83.6	83.8	65.6
Panoptic-PolarNet	63.4	67.2	75.3	83.9	59.2	70.3	84.1	70.4	83.5	83.6	66.9
SCAN^［26］	65.1	68.9	75.3	85.7	60.6	70.2	85.7	72.5	83.8	85.7	77.4
DS-Net^［5］	42.5	51.0	50.3	83.6	32.5	38.3	83.1	59.2	70.3	84.4	70.7
MaskPLS（Baseline）	57.7	60.2	66.0	71.8	64.4	73.3	84.8	52.2	60.7	62.4	62.5
BEVGuide-PS（L）	66.1	69.1	76.8	85.9	65.2	76.5	85.7	72.9	84.1	85.9	77.8
BEVGuide-PS（L+C）	67.9	70.4	78.1	86.8	66.8	77.3	86.3	73.1	84.3	86.5	78.9

Table 2. Comparison of ablation experiment results

View table

Table 2. Comparison of ablation experiment results

Model	BEV object queries fusion	BEV feature-guided	Camera images	PQ /%	PQ^† /%	RQ /%	SQ /%	Inference time /ms
MaskPLS（Baseline）				57.7	60.2	66.0	71.8	315
Model A				52.3	56.1	61.2	69.1	106
Model B	√			56.7	59.2	64.5	70.6	130
BEVGuide-PS（L）	√	√		66.1	69.1	76.8	85.9	131
BEVGuide-PS（L+C）	√	√	√	67.9	70.4	78.1	86.8	163

Tools

Get Citation

Copy Citation Text

Jiaming Zhang, Yuhui Peng, Gan Zhang, Baozhe Sun, Shenyang Lin. Mask-Based and Multi-Modal Bird's Eye View Feature-Guided Point Cloud Panoptic Segmentation[J]. Laser & Optoelectronics Progress, 2025, 62(12): 1215001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Nov. 13, 2024

Accepted: Dec. 12, 2024

Published Online: Jun. 12, 2025

The Author Email: Yuhui Peng (pengyuhui@fzu.edu.cn)

DOI:10.3788/LOP242258

CSTR:32186.14.LOP242258

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology