%0 Conference Proceedings
%T Real-time single-shot brand logo recognition
%D 2017
%8 Oct. 17-20, 2017
%A Bombonato, Leonardo,
%A Camara-Chavez, Guillermo,
%A Silva, Pedro,
%@affiliation Universidade Federal de Ouro Preto
%@affiliation Universidade Federal de Ouro Preto
%@affiliation Universidade Federal de Ouro Preto
%E Torchelsen, Rafael Piccin,
%E Nascimento, Erickson Rangel do,
%E Panozzo, Daniele,
%E Liu, Zicheng,
%E Farias, Mylène,
%E Viera, Thales,
%E Sacht, Leonardo,
%E Ferreira, Nivan,
%E Comba, João Luiz Dihl,
%E Hirata, Nina,
%E Schiavon Porto, Marcelo,
%E Vital, Creto,
%E Pagot, Christian Azambuja,
%E Petronetto, Fabiano,
%E Clua, Esteban,
%E Cardeal, Flávio,
%B Conference on Graphics, Patterns and Images, 30 (SIBGRAPI)
%C Niterói, RJ
%S Proceedings
%I IEEE Computer Society
%J Los Alamitos
%K computer vision, logo recognition, deep learning.
%X The amount of data produced every day on theinternet increases every day and with the increasing popularityof the social networks the number of published photos arehuge, and those pictures contain several implicit or explicitbrand logos. Detecting this logos in natural images can provideinformation about how widespread is a brand, discover unwantedcopyright distribution, analyze marketing campaigns, etc. In thispaper, we propose a real-time brand logo recognition system thatoutperforms all other state-of-the-art in two different datasets.Our approach is based on the Single Shot MultiBox Detector(SSD), we explore this tool in a different domain and alsoexperiment the impact of training with pretrained weights andthe impact of warp transformations in the input images. Weconducted our experiments in two datasets, the FlickrLogos-32(FL32) and the Logos-32Plus (L32plus), which is an extension ofthe training set of the FL32. On the FL32, we outperform thestate-of-the-art by 2.5% the F-score and by 7.4% the recall. Forthe L32plus, we surpass the state-of-the-art by 1.2% the F-scoreand by 3.8% the recall.
%@language pt
%3 Sibgrapi Final Version.pdf