Title:	A new deep convolutional neural network for 2D contour classification
Authors:	Mhedhbi, Makrem Mhiri, Slim Ghorbel, Faouzi
Citation:	WSCG 2022: full papers proceedings: 30. International Conference in Central Europe on Computer Graphics, Visualization and Computer Vision, p. 119-127.
Issue Date:	2022
Publisher:	Václav Skala - UNION Agency
Document type:	conferenceObject
URI:	http://hdl.handle.net/11025/49585
ISBN:	978-80-86943-33-6
ISSN:	2464-4617
Keywords:	popis obrysu;klasifikace tvaru;CNN;hluboké učení;augmentace dat;vývoj křivky
Keywords in different language:	contour description;shape classification;CNN;deep learning;data augmentation;curve evolution
Abstract in different language:	In this paper, we present a new deep convolutional neural network to classify 2d contours, described by a sequence of points coordinates representing the boundary of objects. Several works dealt with this subject, even those using learning, but few works use deep learning. This is due to the fact that contours data are very narrow and inappropriate for convolution. To enrich this representation, we use curve evolution and consider simultaneously a multi-scale representation of a contour. Associated with coordinates, curvature estimated at each point is the most used descriptor who can help distinguishing objects. Despite deficiency of large 2d contour datasets, required for a convergent learning, the use of several additional techniques, such as data augmentation, lead to results outperforming the state of the art. We train ContourNet on MPEG-7 database CE-1 part B, witch achieves 100% for Top-1 accuracy rate on MPEG-7 test set, and 91.78% on Kimia216 dataset.
Rights:	© Václav Skala - UNION Agency
Appears in Collections:	WSCG 2022: Full Papers Proceedings

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.