Presentation 2018-05-25
Efficient Object Detection with Event-Driven camera and its implementation on an FPGA
Masayuki Shimoda, Shimpei Sato, Hiroki Nakahara,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We propose an object detection system using a sliding window method for an event-driven camera which outputs a subtracted frame~(usually binary value) when changes are detected in captured images. Since it skips unchanged frames, it operates faster with lower power consumption compared with a straightforward sliding window approach. Since the event-driven camera output is binary precision frames, an all binarized convolutional neural network~(ABCNN) can be available. Although the realization of all binarization decreases the classification accuracy, it allows all binarized convolutional layers to share the same circuit, thus reducing the area requirement. We implemented our proposed method on a zcu102 FPGA evaluation board and then evaluated it using the Caltech Pedestrian Detection Benchmark. The results show that even though our proposed method reduced a recognition accuracy by 6 points, the computation time for an entire frame was 74.8 times faster than the that of the conventional BCNN without an event-driven camera. Compared with the object detector on the mobile GPU~(NVIDIA Jetson TX2), frames per second~(FPS) of the FPGA system was 4.3 times faster, and the performance per power efficiency was approximately 54.2 times higher.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Event-Driven Camera / Object Detection / CNN / Binarized CNN
Paper # RECONF2018-17
Date of Issue 2018-05-17 (RECONF)

Conference Information
Committee RECONF
Conference Date 2018/5/24(2days)
Place (in Japanese) (See Japanese page)
Place (in English) GATE CITY OHSAKI
Topics (in Japanese) (See Japanese page)
Topics (in English) Deep Learning, Reconfigurable Systems, etc.
Chair Masato Motomura(Hokkaido Univ.)
Vice Chair Yuichiro Shibata(Nagasaki Univ.) / Kentaro Sano(Tohoku Univ.)
Secretary Yuichiro Shibata(Hiroshima City Univ.) / Kentaro Sano(e-trees.Japan)
Assistant Yuuki Kobayashi(NEC) / Hiroki Nakahara(Tokyo Inst. of Tech.)

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) Efficient Object Detection with Event-Driven camera and its implementation on an FPGA
Sub Title (in English)
Keyword(1) Event-Driven Camera
Keyword(2) Object Detection
Keyword(3) CNN
Keyword(4) Binarized CNN
Keyword(5)
1st Author's Name Masayuki Shimoda
1st Author's Affiliation Tokyo Institute of Technology(titech)
2nd Author's Name Shimpei Sato
2nd Author's Affiliation Tokyo Institute of Technology(titech)
3rd Author's Name Hiroki Nakahara
3rd Author's Affiliation Tokyo Institute of Technology(titech)
Date 2018-05-25
Paper # RECONF2018-17
Volume (vol) vol.118
Number (no) RECONF-63
Page pp.pp.81-86(RECONF),
#Pages 6
Date of Issue 2018-05-17 (RECONF)