Co-Training for Deep Object Detection

Top-performing computer vision models are powered by convolutional neural networks (CNNs). Training an accurate CNN highly depends on both the raw sensor data and their associated ground truth (GT). Collecting such GT is usually done through human labeling, which is time-consuming and does not scale...

ver descrição completa

Detalhes bibliográficos
Autores: Gomez Zurita, Jose Luis|||0000-0001-9511-1915, Villalonga, Gabriel|||0000-0002-1155-9374, López Peña, Antonio M.|||0000-0002-6979-5783
Formato: artículo
Fecha de publicación:2021
País:España
Recursos:Universitat Autònoma de Barcelona
Repositorio:Dipòsit Digital de Documents de la UAB
Idioma:inglés
OAI Identifier:oai:ddd.uab.cat:255626
Acesso em linha:https://ddd.uab.cat/record/255626
https://dx.doi.org/urn:doi:10.3390/s21093185
Access Level:acceso abierto
Palavra-chave:Co-training
Multi-modality
Vision-based object detection
ADAS
Self-driving
id ES_19cc2ca608913ffb03c8b997f0e1260b
oai_identifier_str oai:ddd.uab.cat:255626
network_acronym_str ES
network_name_str España
repository_id_str
spelling Co-Training for Deep Object DetectionComparing Single-Modal and Multi-Modal ApproachesGomez Zurita, Jose Luis|||0000-0001-9511-1915Villalonga, Gabriel|||0000-0002-1155-9374López Peña, Antonio M.|||0000-0002-6979-5783Co-trainingMulti-modalityVision-based object detectionADASSelf-drivingTop-performing computer vision models are powered by convolutional neural networks (CNNs). Training an accurate CNN highly depends on both the raw sensor data and their associated ground truth (GT). Collecting such GT is usually done through human labeling, which is time-consuming and does not scale as we wish. This data-labeling bottleneck may be intensified due to domain shifts among image sensors, which could force per-sensor data labeling. In this paper, we focus on the use of co-training, a semi-supervised learning (SSL) method, for obtaining self-labeled object bounding boxes (BBs), i.e., the GT to train deep object detectors. In particular, we assess the goodness of multi-modal co-training by relying on two different views of an image, namely, appearance (RGB) and estimated depth (D). Moreover, we compare appearance-based single-modal co-training with multi-modal. Our results suggest that in a standard SSL setting (no domain shift, a few human-labeled data) and under virtual-to-real domain shift (many virtual-world labeled data, no human-labeled data) multi-modal co-training outperforms single-modal. In the latter case, by performing GAN-based domain translation both co-training modalities are on par, at least when using an off-the-shelf depth estimation model not specifically trained on the translated images. 22021-01-0120212021-01-01Articlehttp://purl.org/coar/resource_type/c_6501VoRhttp://purl.org/coar/version/c_970fb48d4fbd8a85info:eu-repo/semantics/articleapplication/pdfhttps://ddd.uab.cat/record/255626https://dx.doi.org/urn:doi:10.3390/s21093185reponame:Dipòsit Digital de Documents de la UABinstname:Universitat Autònoma de BarcelonaInglésengAgencia Estatal de Investigación https://doi.org/10.13039/501100011033 TIN2017-88709-RMinisterio de Educación, Cultura y Deporte https://doi.org/10.13039/501100003176 FPU16/04131open accesshttp://purl.org/coar/access_right/c_abf2Aquest document està subjecte a una llicència d'ús Creative Commons. Es permet la reproducció total o parcial, la distribució, la comunicació pública de l'obra i la creació d'obres derivades, sempre que no sigui amb finalitats comercials, i sempre que es reconegui l'autoria de l'obra original.https://creativecommons.org/licenses/by-nc/4.0/info:eu-repo/semantics/openAccessoai:ddd.uab.cat:2556262026-06-06T12:50:31Z
dc.title.none.fl_str_mv Co-Training for Deep Object Detection
Comparing Single-Modal and Multi-Modal Approaches
title Co-Training for Deep Object Detection
spellingShingle Co-Training for Deep Object Detection
Gomez Zurita, Jose Luis|||0000-0001-9511-1915
Co-training
Multi-modality
Vision-based object detection
ADAS
Self-driving
title_short Co-Training for Deep Object Detection
title_full Co-Training for Deep Object Detection
title_fullStr Co-Training for Deep Object Detection
title_full_unstemmed Co-Training for Deep Object Detection
title_sort Co-Training for Deep Object Detection
dc.creator.none.fl_str_mv Gomez Zurita, Jose Luis|||0000-0001-9511-1915
Villalonga, Gabriel|||0000-0002-1155-9374
López Peña, Antonio M.|||0000-0002-6979-5783
author Gomez Zurita, Jose Luis|||0000-0001-9511-1915
author_facet Gomez Zurita, Jose Luis|||0000-0001-9511-1915
Villalonga, Gabriel|||0000-0002-1155-9374
López Peña, Antonio M.|||0000-0002-6979-5783
author_role author
author2 Villalonga, Gabriel|||0000-0002-1155-9374
López Peña, Antonio M.|||0000-0002-6979-5783
author2_role author
author
dc.subject.none.fl_str_mv Co-training
Multi-modality
Vision-based object detection
ADAS
Self-driving
topic Co-training
Multi-modality
Vision-based object detection
ADAS
Self-driving
description Top-performing computer vision models are powered by convolutional neural networks (CNNs). Training an accurate CNN highly depends on both the raw sensor data and their associated ground truth (GT). Collecting such GT is usually done through human labeling, which is time-consuming and does not scale as we wish. This data-labeling bottleneck may be intensified due to domain shifts among image sensors, which could force per-sensor data labeling. In this paper, we focus on the use of co-training, a semi-supervised learning (SSL) method, for obtaining self-labeled object bounding boxes (BBs), i.e., the GT to train deep object detectors. In particular, we assess the goodness of multi-modal co-training by relying on two different views of an image, namely, appearance (RGB) and estimated depth (D). Moreover, we compare appearance-based single-modal co-training with multi-modal. Our results suggest that in a standard SSL setting (no domain shift, a few human-labeled data) and under virtual-to-real domain shift (many virtual-world labeled data, no human-labeled data) multi-modal co-training outperforms single-modal. In the latter case, by performing GAN-based domain translation both co-training modalities are on par, at least when using an off-the-shelf depth estimation model not specifically trained on the translated images.
publishDate 2021
dc.date.none.fl_str_mv 2
2021-01-01
2021
2021-01-01
dc.type.none.fl_str_mv Article
http://purl.org/coar/resource_type/c_6501
VoR
http://purl.org/coar/version/c_970fb48d4fbd8a85
dc.type.openaire.fl_str_mv info:eu-repo/semantics/article
format article
dc.identifier.none.fl_str_mv https://ddd.uab.cat/record/255626
https://dx.doi.org/urn:doi:10.3390/s21093185
url https://ddd.uab.cat/record/255626
https://dx.doi.org/urn:doi:10.3390/s21093185
dc.language.none.fl_str_mv Inglés
eng
language_invalid_str_mv Inglés
language eng
dc.relation.none.fl_str_mv Agencia Estatal de Investigación https://doi.org/10.13039/501100011033 TIN2017-88709-R
Ministerio de Educación, Cultura y Deporte https://doi.org/10.13039/501100003176 FPU16/04131
dc.rights.none.fl_str_mv open access
http://purl.org/coar/access_right/c_abf2
https://creativecommons.org/licenses/by-nc/4.0/
dc.rights.openaire.fl_str_mv info:eu-repo/semantics/openAccess
rights_invalid_str_mv open access
http://purl.org/coar/access_right/c_abf2
https://creativecommons.org/licenses/by-nc/4.0/
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.source.none.fl_str_mv reponame:Dipòsit Digital de Documents de la UAB
instname:Universitat Autònoma de Barcelona
instname_str Universitat Autònoma de Barcelona
reponame_str Dipòsit Digital de Documents de la UAB
collection Dipòsit Digital de Documents de la UAB
repository.name.fl_str_mv
repository.mail.fl_str_mv
_version_ 1869404068130062336
score 15.228081

Registos relacionados