Deep learning of appearance affinity for multi-object tracking and re-identification: a comparative view

e-Archivo Repository

Show simple item record

dc.contributor.author Gómez Silva, María José
dc.contributor.author Escalera Hueso, Arturo de la
dc.contributor.author Armingol Moreno, José María
dc.date.accessioned 2021-02-24T13:42:52Z
dc.date.available 2021-02-24T13:42:52Z
dc.date.issued 2020-10-22
dc.identifier.bibliographicCitation Electronics, 9(11), 1757, Oct. 2020, 28 pp.
dc.identifier.issn 2079-9292
dc.identifier.uri http://hdl.handle.net/10016/32014
dc.description.abstract Recognizing the identity of a query individual in a surveillance sequence is the core of Multi-Object Tracking (MOT) and Re-Identification (Re-Id) algorithms. Both tasks can be addressed by measuring the appearance affinity between people observations with a deep neural model. Nevertheless, the differences in their specifications and, consequently, in the characteristics and constraints of the available training data for each one of these tasks, arise from the necessity of employing different learning approaches to attain each one of them. This article offers a comparative view of the Double-Margin-Contrastive and the Triplet loss function, and analyzes the benefits and drawbacks of applying each one of them to learn an Appearance Affinity model for Tracking and Re-Identification. A batch of experiments have been conducted, and their results support the hypothesis concluded from the presented study: Triplet loss function is more effective than the Contrastive one when an Re-Id model is learnt, and, conversely, in the MOT domain, the Contrastive loss can better discriminate between pairs of images rendering the same person or not.
dc.description.sponsorship This research was funded by the Spanish Government through the CICYT projects (TRA2016-78886-C3-1-R and RTI2018-096036-B-C21), Universidad Carlos III of Madrid through (PEAVAUTO-CM-UC3M), the Comunidad de Madrid through SEGVAUTO-4.0-CM (P2018/EMT-4362), and the Ministerio de Educación, Cultura y Deporte para la Formación de Profesorado Universitario (FPU14/02143).
dc.format.extent 28
dc.language.iso eng
dc.publisher MDPI
dc.rights © 2020 by the authors. Licensee MDPI, Basel, Switzerland.
dc.rights This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license
dc.rights Atribución 3.0 España
dc.rights.uri http://creativecommons.org/licenses/by/3.0/es/
dc.subject.other Appearance affinity
dc.subject.other Triplet model
dc.subject.other Contrastive loss function
dc.subject.other Deep convolutional
dc.subject.other Neural network
dc.subject.other Re-identification
dc.subject.other Multi-object tracking
dc.title Deep learning of appearance affinity for multi-object tracking and re-identification: a comparative view
dc.type article
dc.subject.eciencia Robótica e Informática Industrial
dc.identifier.doi https://doi.org/10.3390/electronics9111757
dc.rights.accessRights openAccess
dc.relation.projectID Gobierno de España. TRA2016-78886-C3-1-R
dc.relation.projectID Gobierno de España. RTI2018-096036-B-C21
dc.relation.projectID Gobierno de España. FPU14/02143
dc.relation.projectID Comunidad de Madrid. P2018/EMT-4362-SEGVAUTO-4.0-CM
dc.relation.projectID Comunidad de Madrid. PEAVAUTO-CM-UC3M
dc.type.version publishedVersion
dc.identifier.publicationfirstpage 1
dc.identifier.publicationissue 11, 1757
dc.identifier.publicationlastpage 28
dc.identifier.publicationtitle Electronics
dc.identifier.publicationvolume 9
dc.identifier.uxxi AR/0000026306
dc.contributor.funder Comunidad de Madrid
dc.contributor.funder Ministerio de Economía y Competitividad (España)
dc.contributor.funder Universidad Carlos III de Madrid
dc.contributor.funder Ministerio de Educación, Cultura y Deporte (España)
 Find Full text

Files in this item

*Click on file's image for preview. (Embargoed files's preview is not supported)


The following license files are associated with this item:

This item appears in the following Collection(s)

Show simple item record