notesum.ai

Published at November 22

MSSF: A 4D Radar and Camera Fusion Framework With Multi-Stage Sampling for 3D Object Detection in Autonomous Driving

cs.CV

cs.RO

Released Date: November 22, 2024

Authors: Hongsi Liu¹, Jun Liu¹, Guangfeng Jiang¹, Xin Jin¹

Aff.: ¹not specified

Arxiv: http://arxiv.org/abs/2411.15016v1

Refer to caption

Method	Modality	AP in the Entire Annotated Area (%)				AP in the Driving Corridor (%)				FPS
Method	Modality	Car	Pedestrian	Cyclist	mAP	Car	Pedestrian	Cyclist	mAP	FPS
ImVoxelNet (WACV 2022)[58]	$\mathrm{C}$	19.35	5.62	17.53	14.17	49.52	9.68	28.97	29.39	11.1
PointPillars (CVPR 2019)[2]	$\mathrm{R}$	42.19	39.29	66.66	49.38	71.59	50.67	85.23	69.16	106.4
RadarPillarNet (IEEE T-IM 2023)[50]	$\mathrm{R}$	39.30	35.10	63.63	46.01	71.65	42.80	83.14	65.86	-
CenterPoint (CVPR 2022)[49]	$\mathrm{R}$	35.84	41.03	67.11	47.99	70.65	50.14	85.67	68.82	38.3
VoxelNeXt (CVPR 2023)[24]	$\mathrm{R}$	36.98	42.37	68.15	49.17	70.95	51.85	87.33	70.04	31.6
SMURF (IEEE T-IV 2023)[27]	$\mathrm{R}$	43.31	39.09	71.50	50.97	71.74	50.54	86.87	69.72	-
MSSF-V-R	$\mathrm{R}$	38.28	42.93	69.96	50.39	71.76	52.92	88.93	71.21	24.6
MSSF-PP-R	$\mathrm{R}$	42.17	40.28	65.41	49.29	72.04	51.06	83.09	68.73	104.9
PointAugmenting (CVPR 2021)[34]	$\mathrm{R}+\mathrm{C}$	39.62	44.48	73.70	52.60	71.02	48.59	87.57	69.06	$7.9^{\dagger}$
$\text{Focals Conv}^{\ddagger}$ (CVPR 2022)[46]	$\mathrm{R}+\mathrm{C}$	40.01	48.67	75.42	54.70	71.79	53.41	87.53	70.91	$10.4^{\dagger}$
RCFusion (IEEE T-IM 2023)[50]	$\mathrm{R}+\mathrm{C}$	41.70	38.95	68.31	49.65	71.87	47.50	88.33	69.23	-
LXL (IEEE T-IV 2023)[20]	$\mathrm{R}+\mathrm{C}$	42.33	49.48	77.12	56.31	72.18	58.30	88.31	72.93	$6.1^{*}$
UniBEVFusion (Arxiv 2024) [43]	$\mathrm{R}+\mathrm{C}$	42.22	47.11	72.94	54.09	72.10	57.71	93.29	74.37	-
MSSF-V (Ours)	$\mathrm{R}+\mathrm{C}$	52.53	51.58	75.77	59.96	89.08	66.78	88.10	81.32	10.3
MSSF-PP (Ours)	$\mathrm{R}+\mathrm{C}$	60.96	51.28	77.69	63.31	90.60	60.39	88.35	79.78	13.9
PointPillars (CVPR 2019)[2]	$\mathrm{L}$	68.81	51.26	66.00	62.02	90.84	62.80	85.25	79.63	56.1