Center-Aware 3D Object Detection with Attention Mechanism Based on Roadside LiDAR
نویسندگان
چکیده
Infrastructure 3D Object Detection is a pivotal component of Vehicle-Infrastructure Cooperated Autonomous Driving (VICAD). As turning objects account for high proportion traffic at intersections, anchor-free representation in the bird’s-eye view (BEV) more suitable roadside detection. In this work, we propose CetrRoad, simple yet effective center-aware detector with transformer-based detection head object single LiDAR (Light and Ranging). CetrRoad firstly utilizes voxel-based feature encoder module that voxelizes projects raw point cloud into BEV dense representation, following one-stage center proposal initializes candidates based on top N points target heatmap unnormalized 2D Gaussian. Then, taking attending proposals as query embedding, multi-head self-attention multi-scale deformable cross attention can refine predict bounding boxes different classes moving/parked intersection. Extensive experiments analyses demonstrate our method achieves state-of-the-art performance DAIR-V2X-I benchmark an acceptable training time cost, especially Car Cyclist. also reaches comparable results multi-modal fusion Pedestrian. An ablation study demonstrates input provide denser supervision than purified map attention-based head. Moreover, were able to intuitively observe complex environment, proposed model could produce accurate other compared methods fewer false positives, which helpful downstream VICAD tasks.
منابع مشابه
LIDAR-based 3D Object Perception
This paper describes a LIDAR-based perception system for ground robot mobility, consisting of 3D object detection, classification and tracking. The presented system was demonstrated on-board our autonomous ground vehicle MuCAR-3, enabling it to safely navigate in urban traffic-like scenarios as well as in off-road convoy scenarios. The efficiency of our approach stems from the unique combinatio...
متن کامل3D object detection from roadside data using laser scanners
The detection of objects on a given road path by vehicles equipped with range measurement devices is important to many civilian and military applications such as obstacle avoidance in autonomous navigation systems. In this thesis, we develop a method to detect objects of a specific size lying on a road using an acquisition vehicle equipped with forward looking Light Detection And Range (LiDAR) ...
متن کاملAnomaly-based Web Attack Detection: The Application of Deep Neural Network Seq2Seq With Attention Mechanism
Today, the use of the Internet and Internet sites has been an integrated part of the people’s lives, and most activities and important data are in the Internet websites. Thus, attempts to intrude into these websites have grown exponentially. Intrusion detection systems (IDS) of web attacks are an approach to protect users. But, these systems are suffering from such drawbacks as low accuracy in ...
متن کاملTSSD: Temporal Single-Shot Object Detection Based on Attention-Aware LSTM
Temporal object detection has attracted significant attention, but most popular detection methods can not leverage the rich temporal information in video or robotic vision. Although many different algorithms have been developed for video detection task, real-time online approaches are frequently deficient. In this paper, based on attention mechanism and convolutional long short-term memory (Con...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Sustainability
سال: 2023
ISSN: ['2071-1050']
DOI: https://doi.org/10.3390/su15032628