F-E3D: FPGA-based Acceleration of an Efficient 3D Convolutional Neural Network for Human Action Recognition

被引:37
|
作者
Fan, Hongxiang [1 ]
Luo, Cheng [2 ]
Zeng, Chenglong [3 ]
Ferianc, Martin [1 ]
Que, Zhiqiang [1 ]
Liu, Shuanglong [1 ]
Niu, Xinyu [4 ]
Luk, Wayne [1 ]
机构
[1] Imperial Coll London, Sch Engn, Dept Comp, London, England
[2] Fudan Univ, State Key Lab ASIC & Syst, Shanghai, Peoples R China
[3] Tianjin Univ, Sch Microelect, Tianjin, Peoples R China
[4] Corerain Technol Ltd, Shenzhen, Peoples R China
来源
2019 IEEE 30TH INTERNATIONAL CONFERENCE ON APPLICATION-SPECIFIC SYSTEMS, ARCHITECTURES AND PROCESSORS (ASAP 2019) | 2019年
基金
英国工程与自然科学研究理事会;
关键词
D O I
10.1109/ASAP.2019.00-44
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
Three-dimensional convolutional neural networks (3D CNNs) have demonstrated their outstanding classification accuracy for human action recognition (HAR). However, the large number of computations and parameters in 3D CNNs limits their deployability in real-life applications. To address this challenge, this paper adopts an algorithm-hardware co-design method by proposing an efficient 3D CNN building unit called 3D-1 bottleneck residual block (3D-1 BRB) at the algorithm level, and a corresponding FPGA-based hardware architecture called F-E3D at hardware level. Based on 3D-1 BRB, a novel 3D CNN model called E3DNet is developed, which achieves nearly 37 times reduction in model size and 5% improvement in accuracy compared to standard 3D CNNs on the UCF101 dataset. Together with several hardware optimizations, including 3D fused BRB, online blocking and kernel reuse, the proposed F-E3D is nearly 13 times faster than a previous FPGA design for 3D CNNs, with performance and accuracy comparable to other state-of-the-art 3D CNN models on GPU platforms while requiring only 7% of their energy consumption.
引用
收藏
页码:1 / 8
页数:8
相关论文
共 50 条
  • [21] Human Action Recognition using 3D Convolutional Neural Networks with 3D Motion Cuboids in Surveillance Videos
    Arunnehru, J.
    Chamundeeswari, G.
    Bharathi, S. Prasanna
    INTERNATIONAL CONFERENCE ON ROBOTICS AND SMART MANUFACTURING (ROSMA2018), 2018, 133 : 471 - 477
  • [22] 3D CONVOLUTIONAL NEURAL NETWORK WITH MULTI-MODEL FRAMEWORK FOR ACTION RECOGNITION
    Jing, Longlong
    Ye, Yuancheng
    Yang, Xiaodong
    Tian, Yingli
    2017 24TH IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2017, : 1837 - 1841
  • [23] Study of Human Motion Recognition Algorithm Based on Multichannel 3D Convolutional Neural Network
    Ju, Yang
    COMPLEXITY, 2021, 2021
  • [24] Enhanced 3D Action Recognition Based on Deep Neural Network
    Park, Sungjoo
    Kim, Dongchil
    2022 THIRTEENTH INTERNATIONAL CONFERENCE ON UBIQUITOUS AND FUTURE NETWORKS (ICUFN), 2022, : 470 - 472
  • [25] Study on 3D Action Recognition Based on Deep Neural Network
    Park, Sungjoo
    Kim, Dongchil
    2019 INTERNATIONAL CONFERENCE ON ELECTRONICS, INFORMATION, AND COMMUNICATION (ICEIC), 2019, : 309 - 311
  • [26] 3D convolutional neural network for object recognition: a review
    Rahul Dev Singh
    Ajay Mittal
    Rajesh K. Bhatia
    Multimedia Tools and Applications, 2019, 78 : 15951 - 15995
  • [27] A 3D Tensor Representation of Speech and 3D Convolutional Neural Network for Emotion Recognition
    Mohammad Reza Falahzadeh
    Fardad Farokhi
    Ali Harimi
    Reza Sabbaghi-Nadooshan
    Circuits, Systems, and Signal Processing, 2023, 42 : 4271 - 4291
  • [28] A 3D Tensor Representation of Speech and 3D Convolutional Neural Network for Emotion Recognition
    Falahzadeh, Mohammad Reza
    Farokhi, Fardad
    Harimi, Ali
    Sabbaghi-Nadooshan, Reza
    CIRCUITS SYSTEMS AND SIGNAL PROCESSING, 2023, 42 (07) : 4271 - 4291
  • [29] 3D convolutional neural network for object recognition: a review
    Singh, Rahul Dev
    Mittal, Ajay
    Bhatia, Rajesh K.
    MULTIMEDIA TOOLS AND APPLICATIONS, 2019, 78 (12) : 15951 - 15995
  • [30] 3D Human Motion Synthesis Based on Convolutional Neural Network
    Zhou, Dongsheng
    Feng, Xinzhu
    Yi, Pengfei
    Yang, Xin
    Zhang, Qiang
    Wei, Xiaopeng
    Yang, Deyun
    IEEE ACCESS, 2019, 7 : 66325 - 66335