Implementation of a logo detection system based on deep learning strategies for media impact analysis in social networks

Fuertes Coiras, Daniel (2020). Implementation of a logo detection system based on deep learning strategies for media impact analysis in social networks. Thesis (Master thesis), E.T.S.I. Telecomunicación (UPM).

Description

Title: Implementation of a logo detection system based on deep learning strategies for media impact analysis in social networks
Author/s:
  • Fuertes Coiras, Daniel
Contributor/s:
  • Blanco Adán, Carlos Roberto del
Item Type: Thesis (Master thesis)
Masters title: Teoría de la Señal y Comunicaciones
Date: 2020
Subjects:
Freetext Keywords: Logo detector, YOLO, QMUL-OpenLogo, CutMix, Computer Vision, object detection, Artificial Neural Networks, Convolutional Neural Networks, Deep Learning
Faculty: E.T.S.I. Telecomunicación (UPM)
Department: Señales, Sistemas y Radiocomunicaciones
Creative Commons Licenses: Recognition - No derivative works - Non commercial

Full text

[img]
Preview
PDF - Requires a PDF viewer, such as GSview, Xpdf or Adobe Acrobat Reader
Download (4MB) | Preview

Abstract

Nowadays, social networks are one of the principal focus of attention of millions of people all around the world. Companies are aware of this and try to extract from them valuable information to be applied in the design of a good marketing strategy. The objective of this thesis consists in developing a logo detector that is able to determine the location of any brand logo in an image or video. The main purpose is to determine the media impact of any company in the social media, although it could be useful for other applications such as copyright infraction detection. The developed system will be based on a deep learning strategy focused on object detection that will be combined with CutMix, a state-of-the-art training strategy that improves the learning process. The input of the object detector would be an image or frame from a video and the output, the location of every potential logo inside the image. The develop logo detection system will be tested to check the accuracy as a function of the size and distortion of the images. Finally, it is remarkable that most of the existing logo datasets are excessively small. And the few large ones contain weakly labelled and noisy images that are not suitable. Therefore, a carefully selection and potentially aggregation of databases would be also required.

More information

Item ID: 63132
DC Identifier: http://oa.upm.es/63132/
OAI Identifier: oai:oa.upm.es:63132
Deposited by: Biblioteca ETSI Telecomunicación
Deposited on: 22 Jul 2020 08:02
Last Modified: 22 Jul 2020 08:02
  • Logo InvestigaM (UPM)
  • Logo GEOUP4
  • Logo Open Access
  • Open Access
  • Logo Sherpa/Romeo
    Check whether the anglo-saxon journal in which you have published an article allows you to also publish it under open access.
  • Logo Dulcinea
    Check whether the spanish journal in which you have published an article allows you to also publish it under open access.
  • Logo de Recolecta
  • Logo del Observatorio I+D+i UPM
  • Logo de OpenCourseWare UPM