Skip to main content
ATEE 2023 - OpenConference

Papers - accepted for presentation Proceedings »
Certificate
View File
pdf
539KB

Detection of Melanomas Using Ensembles of Deep Convolutional Neural Networks

Melanoma represents one of the most aggressive and dangerous skin cancers, leading to mortality if not detected and treated in time. To help dermatologists in the early detection of melanoma, recently artificial intelligence techniques have been developed and systems based on neural networks capable of detecting these lesions with high precision. The article proposes two implementations of such multi-network systems (assemblies of efficient neural networks) with good performance for melanoma detection from dermatoscopic images. The first model is one based on the fusion of the decisions of several neural networks considering the weights associated with the individual networks. The second model is one of Horizontal Voting based on the voting of some network models obtained from the basic networks, at various numbers of epochs. Both models give relatively good results, the last one having an accuracy of 94.06% in melanoma detection.

Loretta ICHIM
University "Politehnica" of Bucharest
Romania

Razvan-Ionut MITRIC─é
University "Politehnica"of Bucharest
Romania

Dan POPESCU
University "Politehnica"of Bucharest
Romania

 


Powered by OpenConf®
Copyright ©2002-2022 Zakon Group LLC