notesum.ai

Published at November 15

RETR: Multi-View Radar Detection Transformer for Indoor Perception

cs.CV

cs.AI

cs.LG

math.DG

Released Date: November 15, 2024

Authors: Ryoma Yataka¹, Adriano Cardace², Pu Perry Wang, Petros Boufounos¹, Ryuhei Takahashi³

Aff.: ¹Mitsubishi Electric Research Laboratories (MERL), USA; ²Department of Computer Science and Engineering, University of Bologna, Italy; ³Information Technology R&D Center (ITC), Mitsubishi Electric Corporation, Japan

Arxiv: http://arxiv.org/abs/2411.10293v1

$\mathtt{Model}$	$\mathtt{Dim}$	$\mathtt{Input}$	$\mathtt{BBox}\ \mathtt{Loss}$	$\mathtt{AP}$	$\mathtt{AP_{50}}$	$\mathtt{AP_{75}}$	$\mathtt{AR_{1}}$	$\mathtt{AR_{10}}$
RFMask	2D	$\mathtt{H}$ , $\mathtt{V}$	$\mathtt{H}+\mathtt{I}$	31.37	61.50	27.48	33.23	38.41
DETR	2D	$\mathtt{H}$	$\mathtt{H}+\mathtt{I}$	29.38	62.31	25.35	31.32	43.06
DETR (Top- $K$ )	2D	$\mathtt{H}$ , $\mathtt{V}$	$\mathtt{H}+\mathtt{I}$	39.71	82.74	33.29	38.98	52.81
RETR (TPE@Dec.)	3D	$\mathtt{H}$ , $\mathtt{V}$	$\mathtt{H}+\mathtt{V}+\mathtt{I}$	45.94	81.99	44.04	42.03	57.38
RETR	3D	$\mathtt{H}$ , $\mathtt{V}$	$\mathtt{H}+\mathtt{V}+\mathtt{I}$	46.75	83.80	46.06	42.19	57.39