A novel semi-supervised detection approach with weak annotation

Eric K. Tokuda, Gabriel B.A. Ferreira, Claudio Silva, Roberto M. Cesar

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

Abstract

In this work we propose a semi-supervised learning approach for object detection where we use detections from a preexisting detector to train a new detector. We differ from previous works by coming up with a relative quality metric which involves simpler labeling and by proposing a full framework of automatic generation of improved detectors. To validate our method, we collected a comprehensive dataset of more than two thousand hours of streaming from public traffic cameras that contemplates variations in time, location and weather. We used these data to generate and assess with weak labeling a car detector that outperforms popular detectors on hard situations such as rainy weather and low resolution images. Experimental results are reported, thus corroborating the relevance of the proposed approach.

Original languageEnglish (US)
Title of host publication2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages129-132
Number of pages4
Volume2018-April
ISBN (Electronic)9781538665688
DOIs
StatePublished - Sep 21 2018
Event2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018 - Las Vegas, United States
Duration: Apr 8 2018Apr 10 2018

Other

Other2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018
CountryUnited States
CityLas Vegas
Period4/8/184/10/18

Fingerprint

Detectors
Labeling
Supervised learning
Image resolution
Railroad cars
Cameras

ASJC Scopus subject areas

  • Software
  • Computer Vision and Pattern Recognition
  • Computer Science Applications

Cite this

Tokuda, E. K., Ferreira, G. B. A., Silva, C., & Cesar, R. M. (2018). A novel semi-supervised detection approach with weak annotation. In 2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018 - Proceedings (Vol. 2018-April, pp. 129-132). [8470307] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SSIAI.2018.8470307

A novel semi-supervised detection approach with weak annotation. / Tokuda, Eric K.; Ferreira, Gabriel B.A.; Silva, Claudio; Cesar, Roberto M.

2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018 - Proceedings. Vol. 2018-April Institute of Electrical and Electronics Engineers Inc., 2018. p. 129-132 8470307.

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

Tokuda, EK, Ferreira, GBA, Silva, C & Cesar, RM 2018, A novel semi-supervised detection approach with weak annotation. in 2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018 - Proceedings. vol. 2018-April, 8470307, Institute of Electrical and Electronics Engineers Inc., pp. 129-132, 2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018, Las Vegas, United States, 4/8/18. https://doi.org/10.1109/SSIAI.2018.8470307
Tokuda EK, Ferreira GBA, Silva C, Cesar RM. A novel semi-supervised detection approach with weak annotation. In 2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018 - Proceedings. Vol. 2018-April. Institute of Electrical and Electronics Engineers Inc. 2018. p. 129-132. 8470307 https://doi.org/10.1109/SSIAI.2018.8470307
Tokuda, Eric K. ; Ferreira, Gabriel B.A. ; Silva, Claudio ; Cesar, Roberto M. / A novel semi-supervised detection approach with weak annotation. 2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018 - Proceedings. Vol. 2018-April Institute of Electrical and Electronics Engineers Inc., 2018. pp. 129-132
@inproceedings{9f2985db8c244cb69b73d04edb6a0557,
title = "A novel semi-supervised detection approach with weak annotation",
abstract = "In this work we propose a semi-supervised learning approach for object detection where we use detections from a preexisting detector to train a new detector. We differ from previous works by coming up with a relative quality metric which involves simpler labeling and by proposing a full framework of automatic generation of improved detectors. To validate our method, we collected a comprehensive dataset of more than two thousand hours of streaming from public traffic cameras that contemplates variations in time, location and weather. We used these data to generate and assess with weak labeling a car detector that outperforms popular detectors on hard situations such as rainy weather and low resolution images. Experimental results are reported, thus corroborating the relevance of the proposed approach.",
author = "Tokuda, {Eric K.} and Ferreira, {Gabriel B.A.} and Claudio Silva and Cesar, {Roberto M.}",
year = "2018",
month = "9",
day = "21",
doi = "10.1109/SSIAI.2018.8470307",
language = "English (US)",
volume = "2018-April",
pages = "129--132",
booktitle = "2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018 - Proceedings",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - GEN

T1 - A novel semi-supervised detection approach with weak annotation

AU - Tokuda, Eric K.

AU - Ferreira, Gabriel B.A.

AU - Silva, Claudio

AU - Cesar, Roberto M.

PY - 2018/9/21

Y1 - 2018/9/21

N2 - In this work we propose a semi-supervised learning approach for object detection where we use detections from a preexisting detector to train a new detector. We differ from previous works by coming up with a relative quality metric which involves simpler labeling and by proposing a full framework of automatic generation of improved detectors. To validate our method, we collected a comprehensive dataset of more than two thousand hours of streaming from public traffic cameras that contemplates variations in time, location and weather. We used these data to generate and assess with weak labeling a car detector that outperforms popular detectors on hard situations such as rainy weather and low resolution images. Experimental results are reported, thus corroborating the relevance of the proposed approach.

AB - In this work we propose a semi-supervised learning approach for object detection where we use detections from a preexisting detector to train a new detector. We differ from previous works by coming up with a relative quality metric which involves simpler labeling and by proposing a full framework of automatic generation of improved detectors. To validate our method, we collected a comprehensive dataset of more than two thousand hours of streaming from public traffic cameras that contemplates variations in time, location and weather. We used these data to generate and assess with weak labeling a car detector that outperforms popular detectors on hard situations such as rainy weather and low resolution images. Experimental results are reported, thus corroborating the relevance of the proposed approach.

UR - http://www.scopus.com/inward/record.url?scp=85055537347&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85055537347&partnerID=8YFLogxK

U2 - 10.1109/SSIAI.2018.8470307

DO - 10.1109/SSIAI.2018.8470307

M3 - Conference contribution

AN - SCOPUS:85055537347

VL - 2018-April

SP - 129

EP - 132

BT - 2018 IEEE Southwest Symposium on Image Analysis and Interpretation, SSIAI 2018 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

ER -