Presentation 2022-06-08
Consideration of speeding up AI inference processing by cooperative operation of hardware and software
Tomoya Kawakami, Chikako Nakanishi,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) When cooperative processing of AI inference processing between software and hardware, it is difficult to analyze network structures in the Python language, so cooperative processing is not possible. Therefore, we extracted the network information from the learned model and executed it in the C++ language to achieve cooperative operation on a SoCFPGA board and speed up the process. EfficientNetB0 was selected as the target network, and a circuit that enables general-purpose parallel high-speed processing was designed for the convolutional layer (Conv2D layer), which accounts for approximately 90% of the processing time on the CPU.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) SoC FPGA / higher synthesis / EfficientNet / convolutional operation / Ultra96-V2
Paper # RECONF2022-14
Date of Issue 2022-05-31 (RECONF)

Conference Information
Committee RECONF
Conference Date 2022/6/7(2days)
Place (in Japanese) (See Japanese page)
Place (in English) CCS, Univ. of Tsukuba
Topics (in Japanese) (See Japanese page)
Topics (in English) Reconfigurable system, etc.
Chair Kentaro Sano(RIKEN)
Vice Chair Yoshiki Yamaguchi(Tsukuba Univ.) / Tomonori Izumi(Ritsumeikan Univ.)
Secretary Yoshiki Yamaguchi(NEC) / Tomonori Izumi(Toyohashi Univ. of Tech.)
Assistant Yukitaka Takemura(INTEL) / Yasunori Osana(Ryukyu Univ.)

Paper Information
Registration To Technical Committee on Reconfigurable Systems
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Consideration of speeding up AI inference processing by cooperative operation of hardware and software
Sub Title (in English)
Keyword(1) SoC FPGA
Keyword(2) higher synthesis
Keyword(3) EfficientNet
Keyword(4) convolutional operation
Keyword(5) Ultra96-V2
1st Author's Name Tomoya Kawakami
1st Author's Affiliation Osaka Institute of Technology(OIT)
2nd Author's Name Chikako Nakanishi
2nd Author's Affiliation Osaka Institute of Technology(OIT)
Date 2022-06-08
Paper # RECONF2022-14
Volume (vol) vol.122
Number (no) RECONF-60
Page pp.pp.57-62(RECONF),
#Pages 6
Date of Issue 2022-05-31 (RECONF)