Continual 3D Convolutional Neural Networks for Real-time Processing of Videos

被引:3
|
作者
Hedegaard, Lukas [1 ]
Iosifidis, Alexandros [1 ]
机构
[1] Aarhus Univ, Dept Elect & Comp Engn, Aarhus, Denmark
来源
关键词
3D CNN; Human activity recognition; Efficient; Stream processing; Online inference; Continual inference network;
D O I
10.1007/978-3-031-19772-7_22
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We introduce Continual 3D Convolutional Neural Networks (Co3D CNNs), a new computational formulation of spatio-temporal 3D CNNs, in which videos are processed frame-by-frame rather than by clip. In online tasks demanding frame-wise predictions, Co3D CNNs dispense with the computational redundancies of regular 3D CNNs, namely the repeated convolutions over frames, which appear in overlapping clips. We show that Continual 3D CNNs can reuse preexisting 3D-CNN weights to reduce the per-prediction floating point operations (FLOPs) in proportion to the temporal receptive field while retaining similar memory requirements and accuracy. This is validated with multiple models on Kinetics-400 and Charades with remarkable results: CoX3D models attain state-of-the-art complexity/accuracy trade-offs on Kinetics-400 with 12.1-15.3x reductions of FLOPs and 2.3-3.8% improvements in accuracy compared to regular X3D models while reducing peak memory consumption by up to 48%. Moreover, we investigate the transient response of Co3D CNNs at start-up and perform extensive benchmarks of on-hardware processing characteristics for publicly available 3D CNNs.
引用
收藏
页码:369 / 385
页数:17
相关论文
共 50 条
  • [1] Real-Time Detection of Events in Soccer Videos using 3D Convolutional Neural Networks
    Rongved, Olav A. Norgard
    Hicks, Steven A.
    Thambawita, Vajira
    Stensland, Hakon K.
    Zouganeli, Evi
    Johansen, Dag
    Riegler, Michael A.
    Halvorsen, Pal
    2020 IEEE INTERNATIONAL SYMPOSIUM ON MULTIMEDIA (ISM 2020), 2020, : 135 - 144
  • [2] Real-Time 3D Hand Pose Estimation with 3D Convolutional Neural Networks
    Ge, Liuhao
    Liang, Hui
    Yuan, Junsong
    Thalmann, Daniel
    IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2019, 41 (04) : 956 - 970
  • [3] Using 3D Convolutional Neural Networks for Real-time Detection of Soccer Events
    Rongved, Olav A. Nergard
    Hicks, Steven A.
    Thambawita, Vajira
    Stensland, Hakon K.
    Zouganeli, Evi
    Johansen, Dag
    Midoglu, Cise
    Riegler, Michael A.
    Halvorsen, Pal
    INTERNATIONAL JOURNAL OF SEMANTIC COMPUTING, 2021, 15 (02) : 161 - 187
  • [4] RT3D: Achieving Real-Time Execution of 3D Convolutional Neural Networks on Mobile Devices
    Niu, Wei
    Sun, Mengshu
    Li, Zhengang
    Chen, Jou-An
    Guan, Jiexiong
    Shen, Xipeng
    Wang, Yanzhi
    Liu, Sijia
    Lin, Xue
    Ren, Bin
    THIRTY-FIFTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THIRTY-THIRD CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE AND THE ELEVENTH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2021, 35 : 9179 - 9187
  • [5] Real-time 3D Scene Layout from a Single Image Using Convolutional Neural Networks
    Yang, Shichao
    Maturana, Daniel
    Scherer, Sebastian
    2016 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA), 2016, : 2183 - 2189
  • [6] VoxNet: A 3D Convolutional Neural Network for Real-Time Object Recognition
    Maturana, Daniel
    Scherer, Sebastian
    2015 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS), 2015, : 922 - 928
  • [7] A 3D Convolutional Neural Network Towards Real-time Amodal 3D Object Detection
    Sun, Hao
    Meng, Zehui
    Du, Xinxin
    Ang, Marcelo H., Jr.
    2018 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS), 2018, : 8331 - 8338
  • [8] PointNet: A 3D Convolutional Neural Network for Real-Time Object Class Recognition
    Garcia-Garcia, A.
    Gomez-Donoso, F.
    Garcia-Rodriguez, J.
    Orts-Escolano, S.
    Cazorla, M.
    Azorin-Lopez, J.
    2016 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2016, : 1578 - 1584
  • [9] Towards real-time photorealistic 3D holography with deep neural networks
    Shi, Liang
    Li, Beichen
    Kim, Changil
    Kellnhofer, Petr
    Matusik, Wojciech
    NATURE, 2021, 591 (7849) : 234 - +
  • [10] Towards real-time photorealistic 3D holography with deep neural networks
    Liang Shi
    Beichen Li
    Changil Kim
    Petr Kellnhofer
    Wojciech Matusik
    Nature, 2021, 591 : 234 - 239