Exploring Resource-Efficient Acceleration Algorithm for Transposed Convolution of GANs on FPGA

被引:6
|
作者
Di, Xinkai [1 ,2 ]
Yang, Haigang [1 ,2 ]
Huang, Zhihong [1 ,2 ]
Mao, Ning [1 ,2 ]
Jia, Yiping [1 ,2 ]
Zheng, Yong [1 ,2 ]
机构
[1] Univ Chinese Acad Sci, Beijing 100049, Peoples R China
[2] Chinese Acad Sci, Inst Elect, Beijing 100190, Peoples R China
来源
2019 INTERNATIONAL CONFERENCE ON FIELD-PROGRAMMABLE TECHNOLOGY (ICFPT 2019) | 2019年
基金
中国国家自然科学基金;
关键词
Generative Adversarial Networks (GANs); Transposed Convolution; FPGA; Winograd; Hardware Accelerator Architecture;
D O I
10.1109/ICFPT47387.2019.00011
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
In recent years, Generative Adversarial Networks (GANs) have been widely adopted for computer vision tasks such as generation/synthesis of massive images and 3D object modeling. The hardware acceleration of Transposed Convolution layers is especially essential since the Generative Model (Generator) as a critical component in GANs is computationally intensive in nature. In transposed Convolution, the zeros-inserting preprocessing causes sparsity of the feature maps and further results in many invalid operations. Most of the existing FPGA architectures cannot effectively tackle this issue. To address the challenges of implementing Transposed Convolution on FPGAs, we present an innovative dataflow design approach by applying the Winograd algorithm for fast processing with a high efficiency in terms of resource allocations. In addition, we propose an underlying Hardware Accelerator Architecture that features having PUs embedded in Parallel, Pipelined, and Buffered processing flow. In this paper, a parallelism-aware Memory Partition scheme is also exploded for bandwidth efficient data access. Implementations of several state-of-the-art GANs by our approach achieves an average performance of 639.2 GOPS on Xilinx ZCU102 FPGA device. In reference to an optimized conventional accelerator baseline, this work demonstrates an 8.6x (up to 11.7x) improvement in processing performance, compared to below 2.2x improvement by the other works in literature.
引用
收藏
页码:19 / 27
页数:9
相关论文
共 50 条
  • [1] Exploring Efficient Acceleration Architecture for Winograd-Transformed Transposed Convolution of GANs on FPGAs
    Di, Xinkai
    Yang, Hai-Gang
    Jia, Yiping
    Huang, Zhihong
    Mao, Ning
    ELECTRONICS, 2020, 9 (02)
  • [2] Resource-Efficient Optimization for FPGA-Based Convolution Accelerator
    Ma, Yanhua
    Xu, Qican
    Song, Zerui
    ELECTRONICS, 2023, 12 (20)
  • [3] An Intermediate-Centric Dataflow for Transposed Convolution Acceleration on FPGA
    Ma, Zhengzheng
    Dai, Tuo
    Wei, Xuechao
    Luo, Guojie
    ACM TRANSACTIONS ON EMBEDDED COMPUTING SYSTEMS, 2023, 22 (06)
  • [4] HybriDC: A Resource-Efficient CPU-FPGA Heterogeneous Acceleration System for Lossless Data Compression
    Liu, Puguang
    Wei, Ziling
    Yu, Chuan
    Chen, Shuhui
    MICROMACHINES, 2022, 13 (11)
  • [5] Optimizing Deep Learning Acceleration on FPGA for Real-Time and Resource-Efficient Image Classification
    Khaki, Ahmad Mouri Zadeh
    Choi, Ahyoung
    APPLIED SCIENCES-BASEL, 2025, 15 (01):
  • [6] Resource-Efficient FPGA Architecture of Canny Edge Detector
    Jang, Yunseok
    Mun, Junwon
    Kim, Jaeseok
    2016 INTERNATIONAL SOC DESIGN CONFERENCE (ISOCC), 2016, : 299 - 300
  • [7] Resource-Efficient FPGA Implementation of Advanced Encryption Standard
    Lee, Useok
    Kim, Ho Keun
    Lim, Young Jun
    Sunwoo, Myung Hoon
    2022 IEEE INTERNATIONAL SYMPOSIUM ON CIRCUITS AND SYSTEMS (ISCAS 22), 2022, : 1165 - 1169
  • [8] Resource-Efficient FPGA Architecture and Implementation of Hough Transform
    Chen, Zhong-Ho
    Su, Alvin W. Y.
    Sun, Ming-Ting
    IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS, 2012, 20 (08) : 1419 - 1428
  • [9] Resource-efficient quantum algorithm for protein folding
    Anton Robert
    Panagiotis Kl. Barkoutsos
    Stefan Woerner
    Ivano Tavernelli
    npj Quantum Information, 7
  • [10] Resource-efficient quantum algorithm for protein folding
    Robert, Anton
    Barkoutsos, Panagiotis Kl.
    Woerner, Stefan
    Tavernelli, Ivano
    NPJ QUANTUM INFORMATION, 2021, 7 (01)