A sparse deep feature representation for object detection from wearable cameras

Quanfu Fan, Chun Fu Chen, Gwo Giun Lee

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

We propose a novel sparse feature representation for the faster RCNN framework and apply it for object detection from wearable cameras. Two main ideas, sparse convolution and sparse ROI pooling, are developed to reduce model complexity as well as computational cost. Sparse convolution approximates a full kernel by skipping weights in the kernel while sparse ROI pooling performs feature dimensionality reduction on the ROI pooling layer by skipping odd-indexed or even-indexed features. We demonstrate the effectiveness of our approach on two challenging body camera datasets including realistic police-generated clips. Our approach achieves a significant reduction of model size by a factor of over 10× as well as a computational speedup of about 2×, yet without compromising much detection accuracy compared to a VGG16-based baseline detector.

Original languageEnglish
Title of host publicationBritish Machine Vision Conference 2017, BMVC 2017
PublisherBMVA Press
ISBN (Electronic)190172560X, 9781901725605
Publication statusPublished - 2017 Jan 1
Event28th British Machine Vision Conference, BMVC 2017 - London, United Kingdom
Duration: 2017 Sep 42017 Sep 7

Publication series

NameBritish Machine Vision Conference 2017, BMVC 2017

Conference

Conference28th British Machine Vision Conference, BMVC 2017
CountryUnited Kingdom
CityLondon
Period17-09-0417-09-07

Fingerprint

Convolution
Cameras
Law enforcement
Computational complexity
Detectors
Costs
Object detection

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition

Cite this

Fan, Q., Chen, C. F., & Lee, G. G. (2017). A sparse deep feature representation for object detection from wearable cameras. In British Machine Vision Conference 2017, BMVC 2017 (British Machine Vision Conference 2017, BMVC 2017). BMVA Press.
Fan, Quanfu ; Chen, Chun Fu ; Lee, Gwo Giun. / A sparse deep feature representation for object detection from wearable cameras. British Machine Vision Conference 2017, BMVC 2017. BMVA Press, 2017. (British Machine Vision Conference 2017, BMVC 2017).
@inproceedings{00028ff7bca14a64bfafab14d81b1949,
title = "A sparse deep feature representation for object detection from wearable cameras",
abstract = "We propose a novel sparse feature representation for the faster RCNN framework and apply it for object detection from wearable cameras. Two main ideas, sparse convolution and sparse ROI pooling, are developed to reduce model complexity as well as computational cost. Sparse convolution approximates a full kernel by skipping weights in the kernel while sparse ROI pooling performs feature dimensionality reduction on the ROI pooling layer by skipping odd-indexed or even-indexed features. We demonstrate the effectiveness of our approach on two challenging body camera datasets including realistic police-generated clips. Our approach achieves a significant reduction of model size by a factor of over 10× as well as a computational speedup of about 2×, yet without compromising much detection accuracy compared to a VGG16-based baseline detector.",
author = "Quanfu Fan and Chen, {Chun Fu} and Lee, {Gwo Giun}",
year = "2017",
month = "1",
day = "1",
language = "English",
series = "British Machine Vision Conference 2017, BMVC 2017",
publisher = "BMVA Press",
booktitle = "British Machine Vision Conference 2017, BMVC 2017",

}

Fan, Q, Chen, CF & Lee, GG 2017, A sparse deep feature representation for object detection from wearable cameras. in British Machine Vision Conference 2017, BMVC 2017. British Machine Vision Conference 2017, BMVC 2017, BMVA Press, 28th British Machine Vision Conference, BMVC 2017, London, United Kingdom, 17-09-04.

A sparse deep feature representation for object detection from wearable cameras. / Fan, Quanfu; Chen, Chun Fu; Lee, Gwo Giun.

British Machine Vision Conference 2017, BMVC 2017. BMVA Press, 2017. (British Machine Vision Conference 2017, BMVC 2017).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - A sparse deep feature representation for object detection from wearable cameras

AU - Fan, Quanfu

AU - Chen, Chun Fu

AU - Lee, Gwo Giun

PY - 2017/1/1

Y1 - 2017/1/1

N2 - We propose a novel sparse feature representation for the faster RCNN framework and apply it for object detection from wearable cameras. Two main ideas, sparse convolution and sparse ROI pooling, are developed to reduce model complexity as well as computational cost. Sparse convolution approximates a full kernel by skipping weights in the kernel while sparse ROI pooling performs feature dimensionality reduction on the ROI pooling layer by skipping odd-indexed or even-indexed features. We demonstrate the effectiveness of our approach on two challenging body camera datasets including realistic police-generated clips. Our approach achieves a significant reduction of model size by a factor of over 10× as well as a computational speedup of about 2×, yet without compromising much detection accuracy compared to a VGG16-based baseline detector.

AB - We propose a novel sparse feature representation for the faster RCNN framework and apply it for object detection from wearable cameras. Two main ideas, sparse convolution and sparse ROI pooling, are developed to reduce model complexity as well as computational cost. Sparse convolution approximates a full kernel by skipping weights in the kernel while sparse ROI pooling performs feature dimensionality reduction on the ROI pooling layer by skipping odd-indexed or even-indexed features. We demonstrate the effectiveness of our approach on two challenging body camera datasets including realistic police-generated clips. Our approach achieves a significant reduction of model size by a factor of over 10× as well as a computational speedup of about 2×, yet without compromising much detection accuracy compared to a VGG16-based baseline detector.

UR - http://www.scopus.com/inward/record.url?scp=85071159656&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071159656&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:85071159656

T3 - British Machine Vision Conference 2017, BMVC 2017

BT - British Machine Vision Conference 2017, BMVC 2017

PB - BMVA Press

ER -

Fan Q, Chen CF, Lee GG. A sparse deep feature representation for object detection from wearable cameras. In British Machine Vision Conference 2017, BMVC 2017. BMVA Press. 2017. (British Machine Vision Conference 2017, BMVC 2017).