MonoATT: Online Monocular 3D Object Detection with Adaptive Token Transformer

被引:11
|
作者
Zhou, Yunsong [1 ]
Zhu, Hongzi [1 ]
Liu, Quan [1 ]
Chang, Shan [2 ]
Guo, Minyi [1 ]
机构
[1] Shanghai Jiao Tong Univ, Shanghai, Peoples R China
[2] Donghua Univ, Shanghai, Peoples R China
基金
上海市自然科学基金; 中国国家自然科学基金;
关键词
D O I
10.1109/CVPR52729.2023.01678
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Mobile monocular 3D object detection (Mono3D) (e.g., on a vehicle, a drone, or a robot) is an important yet challenging task. Existing transformer-based offline Mono3D models adopt grid-based vision tokens, which is suboptimal when using coarse tokens due to the limited available computational power. In this paper, we propose an online Mono3D framework, called MonoATT, which leverages a novel vision transformer with heterogeneous tokens of varying shapes and sizes to facilitate mobile Mono3D. The core idea of MonoATT is to adaptively assign finer tokens to areas of more significance before utilizing a transformer to enhance Mono3D. To this end, we first use prior knowledge to design a scoring network for selecting the most important areas of the image, and then propose a token clustering and merging network with an attention mechanism to gradually merge tokens around the selected areas in multiple stages. Finally, a pixel-level feature map is reconstructed from heterogeneous tokens before employing a SOTA Mono3D detector as the underlying detection core. Experiment results on the real-world KITTI dataset demonstrate that MonoATT can effectively improve the Mono3D accuracy for both near and far objects and guarantee low latency. MonoATT yields the best performance compared with the state-of-the-art methods by a large margin and is ranked number one on the KITTI 3D benchmark.
引用
收藏
页码:17493 / 17503
页数:11
相关论文
共 50 条
  • [21] Progressive Coordinate Transforms for Monocular 3D Object Detection
    Wang, Li
    Zhang, Li
    Zhu, Yi
    Zhang, Zhi
    He, Tong
    Li, Mu
    Xue, Xiangyang
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 34 (NEURIPS 2021), 2021, 34
  • [22] Exploring Geometric Consistency for Monocular 3D Object Detection
    Lian, Qing
    Ye, Botao
    Xu, Ruijia
    Yao, Weilong
    Zhang, Tong
    2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022), 2022, : 1675 - 1684
  • [23] MonoSG: Monocular 3D Object Detection With Stereo Guidance
    Fan, Zhiwei
    Xu, Chao
    Chu, Minghang
    Huang, Yuling
    Ma, Yaoyao
    Wang, Jing
    Xu, Yishen
    Wu, Di
    IEEE ROBOTICS AND AUTOMATION LETTERS, 2025, 10 (04): : 3604 - 3611
  • [24] Monocular 3D Object Detection With Motion Feature Distillation
    Hu, Henan
    Li, Muyu
    Zhu, Ming
    Gao, Wen
    Liu, Peiyu
    Chan, Kwok-Leung
    IEEE ACCESS, 2023, 11 : 82933 - 82945
  • [25] Monocular Object Detection Using 3D Geometric Primitives
    Carr, Peter
    Sheikh, Yaser
    Matthews, Iain
    COMPUTER VISION - ECCV 2012, PT I, 2012, 7572 : 864 - 878
  • [26] Monocular 3D Object Detection from Roadside Infrastructure
    Huang, Delu
    Wen, Feng
    2024 35TH IEEE INTELLIGENT VEHICLES SYMPOSIUM, IEEE IV 2024, 2024, : 1672 - 1677
  • [27] Dense-JANet for Monocular 3D Object Detection
    Shang, Xiaoqing
    Cheng, Zhiwei
    Shi, Su
    Cheng, Zhuanghao
    Huang, Hongcheng
    2020 IEEE 23RD INTERNATIONAL CONFERENCE ON INTELLIGENT TRANSPORTATION SYSTEMS (ITSC), 2020,
  • [28] DST3D: DLA-Swin Transformer for Single-Stage Monocular 3D Object Detection
    Wu, Zhihong
    Jiang, Xin
    Xu, Ruidong
    Lu, Ke
    Zhu, Yuan
    Wu, Mingzhi
    2022 IEEE INTELLIGENT VEHICLES SYMPOSIUM (IV), 2022, : 411 - 418
  • [29] Monocular 3D object detection for an indoor robot environment
    Kim, Jiwon
    Lee, GiJae
    Kim, Jun-Sik
    Kim, Hyunwoo J.
    Kim, KangGeon
    2020 29TH IEEE INTERNATIONAL CONFERENCE ON ROBOT AND HUMAN INTERACTIVE COMMUNICATION (RO-MAN), 2020, : 438 - 445
  • [30] MonoCD: Monocular 3D Object Detection with Complementary Depths
    Yan, Longfei
    Yan, Pei
    Xiong, Shengzhou
    Xiang, Xuanyu
    Tan, Yihua
    2024 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2024, : 10248 - 10257