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
  • [41] Convolutional and Recurrent Neural Networks for Real-time Data Classification
    Abroyan, Narek
    2017 SEVENTH INTERNATIONAL CONFERENCE ON INNOVATIVE COMPUTING TECHNOLOGY (INTECH 2017), 2017, : 42 - 45
  • [42] Real-time arrhythmia detection using convolutional neural networks
    Vu, Thong
    Petty, Tyler
    Yakut, Kemal
    Usman, Muhammad
    Xue, Wei
    Haas, Francis M.
    Hirsh, Robert A.
    Zhao, Xinghui
    FRONTIERS IN BIG DATA, 2023, 6
  • [43] Real-Time Pedestrian Detection Using Convolutional Neural Networks
    Kuang, Ping
    Ma, Tingsong
    Li, Fan
    Chen, Ziwei
    INTERNATIONAL JOURNAL OF PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE, 2018, 32 (11)
  • [44] Real-Time Grasp Detection Using Convolutional Neural Networks
    Redmon, Joseph
    Angelova, Anelia
    2015 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA), 2015, : 1316 - 1322
  • [45] Real-time Human Pose Estimation with Convolutional Neural Networks
    Linna, Marko
    Kannala, Juho
    Rahtu, Esa
    PROCEEDINGS OF THE 13TH INTERNATIONAL JOINT CONFERENCE ON COMPUTER VISION, IMAGING AND COMPUTER GRAPHICS THEORY AND APPLICATIONS (VISIGRAPP 2018), VOL 5: VISAPP, 2018, : 335 - 342
  • [46] Real-Time Query Processing on Live Videos in Networks of Distributed Cameras
    Peng, Rui
    Aved, Alex J.
    Hua, Kien A.
    INTERNATIONAL JOURNAL OF INTERDISCIPLINARY TELECOMMUNICATIONS AND NETWORKING, 2010, 2 (01) : 27 - 48
  • [47] 3D object understanding with 3D Convolutional Neural Networks
    Leng, Biao
    Liu, Yu
    Yu, Kai
    Zhang, Xiangyang
    Xiong, Zhang
    INFORMATION SCIENCES, 2016, 366 : 188 - 201
  • [48] Real-time monitoring of driver drowsiness on mobile platforms using 3D neural networks
    Wijnands, Jasper S.
    Thompson, Jason
    Nice, Kerry A.
    Aschwanden, Gideon D. P. A.
    Stevenson, Mark
    NEURAL COMPUTING & APPLICATIONS, 2020, 32 (13): : 9731 - 9743
  • [49] Author Correction: Towards real-time photorealistic 3D holography with deep neural networks
    Liang Shi
    Beichen Li
    Changil Kim
    Petr Kellnhofer
    Wojciech Matusik
    Nature, 2021, 593 : E13 - E13
  • [50] Real-time monitoring of driver drowsiness on mobile platforms using 3D neural networks
    Jasper S. Wijnands
    Jason Thompson
    Kerry A. Nice
    Gideon D. P. A. Aschwanden
    Mark Stevenson
    Neural Computing and Applications, 2020, 32 : 9731 - 9743