Fully Automated DCNN-Based Thermal Images Annotation Using Neural Network Pretrained on RGB Data

Simple item page
dc.contributor.authorLigocki, Adamcs
dc.contributor.authorJelínek, Alešcs
dc.contributor.authorŽalud, Luděkcs
dc.contributor.authorRahtu, Esacs
dc.coverage.issue4cs
dc.coverage.volume21cs
dc.date.accessioned2021-03-11T11:53:33Z
dc.date.available2021-03-11T11:53:33Z
dc.date.issued2021-02-24cs
dc.description.abstractOne of the biggest challenges of training deep neural network is the need for massive data annotation. To train the neural network for object detection, millions of annotated training images are required. However, currently, there are no large-scale thermal image datasets that could be used to train the state of the art neural networks, while voluminous RGB image datasets are available. This paper presents a method that allows to create hundreds of thousands of annotated thermal images using the RGB pre-trained object detector. A dataset created in this way can be used to train object detectors with improved performance. The main gain of this work is the novel method for fully automatic thermal image labeling. The proposed system uses the RGB camera, thermal camera, 3D LiDAR, and the pre-trained neural network that detects objects in the RGB domain. Using this setup, it is possible to run the fully automated process that annotates the thermal images and creates the automatically annotated thermal training dataset. As the result, we created a dataset containing hundreds of thousands of annotated objects. This approach allows to train deep learning models with similar performance as the common human-annotation-based methods do. This paper also proposes several improvements to fine-tune the results with minimal human intervention. Finally, the evaluation of the proposed solution shows that the method gives significantly better results than training the neural network with standard small-scale hand-annotated thermal image datasets.en
dc.formattextcs
dc.format.extent1-23cs
dc.format.mimetypeapplication/pdfcs
dc.identifier.citationSENSORS. 2021, vol. 21, issue 4, p. 1-23.en
dc.identifier.doi10.3390/s21041552cs
dc.identifier.issn1424-8220cs
dc.identifier.other170007cs
dc.identifier.urihttp://hdl.handle.net/11012/196455
dc.language.isoencs
dc.publisherMDPIcs
dc.relation.ispartofSENSORScs
dc.relation.urihttps://www.mdpi.com/1424-8220/21/4/1552cs
dc.rightsCreative Commons Attribution 4.0 Internationalcs
dc.rights.accessopenAccesscs
dc.rights.sherpahttp://www.sherpa.ac.uk/romeo/issn/1424-8220/cs
dc.rights.urihttp://creativecommons.org/licenses/by/4.0/cs
dc.subjectdeep convolutional neural networksen
dc.subjecttransfer learningen
dc.subjectYOLOen
dc.subjectRGBen
dc.subjectIRen
dc.subjectthermalen
dc.subjectdata annotationen
dc.subjectobject detectoren
dc.titleFully Automated DCNN-Based Thermal Images Annotation Using Neural Network Pretrained on RGB Dataen
dc.type.driverarticleen
dc.type.statusPeer-revieweden
dc.type.versionpublishedVersionen
sync.item.dbidVAV-170007en
sync.item.dbtypeVAVen
sync.item.insts2021.03.26 12:54:05en
sync.item.modts2021.03.26 12:14:33en
thesis.grantorVysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií. Ústav automatizace a měřicí technikycs
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
sensors2101552v2.pdf
Size:
20.24 MB
Format:
Adobe Portable Document Format
Description:
sensors2101552v2.pdf
Download
Collections
Ústav automatizace a měřicí techniky