计算机工程与应用2024，Vol.60Issue(19)：178-189,12.DOI:10.3778/j.issn.1002-8331.2307-0084
基于Contextual Transformer的自动驾驶单目3D目标检测
Monocular 3D Object Detection for Autonomous Driving Based on Contextual Transformer
摘要
Abstract
Aiming at the current problems of leakage and poor multi-scale target detection in monocular 3D object detec-tion,a monocular 3D object detection algorithm for autonomous driving based on Contextual Transformer(CM-RTM3D)is proposed.Firstly,Contextual Transformer(CoT)is introduced into the ResNet-50 network to construct the ResNet-Transformer architecture for feature extraction.Secondly,the multi-scale spatial perception(MSP)module is designed to improve the loss of shallow features through scale-space response operations,embedding the coordinate attention mecha-nism(CA)along both horizontal and vertical spatial directions,and generating soft weights of importance at each scale using the softmax function.Finally,the Huber loss function is used instead of the L1 loss function in the offset loss.The experi-mental results show that,compared with the RTM3D algorithm on the KITTI autopilot dataset,the algorithm in this paper improves AP3D by 4.84,3.82,and 5.36 percentage points,and APBEV by 4.75,6.26,and 3.56 percentage points,respectively,at the three difficulty levels of easy,medium,and difficult.关键词
自动驾驶/单目3D目标检测/Contextual Transformer/多尺度感知/坐标注意力机制Key words
autonomous driving/monocular 3D object detection/Contextual Transformer/multi-scale perception/coordi-nate attention mechanism分类
计算机与自动化引用本文复制引用
厍向阳,颜唯佳,董立红..基于Contextual Transformer的自动驾驶单目3D目标检测[J].计算机工程与应用,2024,60(19):178-189,12.基金项目
陕西省自然科学基础研究项目(2019JLM-11) （2019JLM-11）
陕西省科技计划(2021JQ-576) （2021JQ-576）
陕西省教育厅项目(19JK0526). （19JK0526）
