LidarMultiNet: Towards a Unified Multi-Task Network for LiDAR Perception

被引:0
|
作者
Ye, Dongqiangzi [1 ]
Zhou, Zixiang [1 ,2 ]
Chen, Weijia [1 ]
Xie, Yufei [1 ]
Wang, Yu [1 ]
Wang, Panqu [1 ]
Foroosh, Hassan [2 ]
机构
[1] TuSimple, San Diego, CA 92122 USA
[2] Univ Cent Florida, Orlando, FL USA
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
LiDAR-based 3D object detection, semantic segmentation, and panoptic segmentation are usually implemented in specialized networks with distinctive architectures that are difficult to adapt to each other. This paper presents LidarMultiNet, a LiDAR-based multi-task network that unifies these three major LiDAR perception tasks. Among its many benefits, a multi-task network can reduce the overall cost by sharing weights and computation among multiple tasks. However, it typically underperforms compared to independently combined single-task models. The proposed LidarMultiNet aims to bridge the performance gap between the multi-task network and multiple single-task networks. At the core of LidarMultiNet is a strong 3D voxel-based encoder-decoder architecture with a Global Context Pooling (GCP) module extracting global contextual features from a LiDAR frame. Task-specific heads are added on top of the network to perform the three LiDAR perception tasks. More tasks can be implemented simply by adding new task-specific heads while introducing little additional cost. A second stage is also proposed to refine the first-stage segmentation and generate accurate panoptic segmentation results. LidarMultiNet is extensively tested on both Waymo Open Dataset and nuScenes dataset, demonstrating for the first time that major LiDAR perception tasks can be unified in a single strong network that is trained end-to-end and achieves state-of-the-art performance. Notably, LidarMultiNet reaches the official 1st place in the Waymo Open Dataset 3D semantic segmentation challenge 2022 with the highest mIoU and the best accuracy for most of the 22 classes on the test set, using only LiDAR points as input. It also sets the new state-of-the-art for a single model on the Waymo 3D object detection benchmark and three nuScenes benchmarks.
引用
收藏
页码:3231 / 3240
页数:10
相关论文
共 50 条
  • [1] GDMNet: A Unified Multi-Task Network for Panoptic Driving Perception
    Liu, Yunxiang
    Ma, Haili
    Zhu, Jianlin
    Zhang, Qiangbo
    CMC-COMPUTERS MATERIALS & CONTINUA, 2024, 80 (02): : 2963 - 2978
  • [2] LiDAR-Based Multi-Task Road Perception Network for Autonomous Vehicles
    Yan, Fuwu
    Wang, Kewei
    Zou, Bin
    Tang, Luqi
    Li, Wenbo
    Lv, Chen
    IEEE ACCESS, 2020, 8 : 86753 - 86764
  • [3] CenterPNets: A Multi-Task Shared Network for Traffic Perception
    Chen, Guangqiu
    Wu, Tao
    Duan, Jin
    Hu, Qi
    Huang, Dandan
    Li, Hao
    SENSORS, 2023, 23 (05)
  • [4] LiDAR-BEVMTN: Real-Time LiDAR Bird's-Eye View Multi-Task Perception Network for Autonomous Driving
    Mohapatra, Sambit
    Yogamani, Senthil
    Kumar, Varun Ravi
    Milz, Stefan
    Gotzig, Heinrich
    Maeder, Patrick
    IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, 2025, 26 (02) : 1547 - 1561
  • [5] A LiDAR-Based Dynamic Driving Scene Multi-task Segmentation Network
    Wang, Hai
    Li, Jianguo
    Cai, Yingfeng
    Chen, Long
    Qiche Gongcheng/Automotive Engineering, 2024, 46 (09): : 1608 - 1616
  • [6] RepVF: A Unified Vector Fields Representation for Multi-task 3D Perception
    Li, Chunliang
    Han, Wencheng
    Yin, Junbo
    Zhao, Sanyuan
    Shen, Jianbing
    COMPUTER VISION - ECCV 2024, PT XXXII, 2025, 15090 : 273 - 292
  • [7] Sparse U-PDP: A Unified Multi-Task Framework for Panoptic Driving Perception
    Wang, Hai
    Qiu, Meng
    Cai, Yingfeng
    Chen, Long
    Li, Yicheng
    IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, 2023, 24 (10) : 11308 - 11320
  • [8] MULTI-TASK DISTILLATION: TOWARDS MITIGATING THE NEGATIVE TRANSFER IN MULTI-TASK LEARNING
    Meng, Ze
    Yao, Xin
    Sun, Lifeng
    2021 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2021, : 389 - 393
  • [9] Multi-task Network Embedding
    Xu, Linchuan
    Wei, Xiaokai
    Cao, Jiannong
    Yu, Philip S.
    2017 IEEE INTERNATIONAL CONFERENCE ON DATA SCIENCE AND ADVANCED ANALYTICS (DSAA), 2017, : 571 - 580
  • [10] ConnectomeNet: A Unified Deep Neural Network Modeling Framework for Multi-Task Learning
    Lim, Heechul
    Chon, Kang-Wook
    Kim, Min-Soo
    IEEE ACCESS, 2023, 11 : 34297 - 34308