Open AccessImage synthesis with neural networks for traffic sign classification
Author(s) -
Vlad Shakhuro,
Anton Konushin
Publication year - 2018
Publication title -
kompʹûternaâ optika
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.491
H-Index - 29
eISSN - 2412-6179
pISSN - 0134-2452
DOI - 10.18287/2412-6179-2018-42-1-105-112
Subject(s) - traffic sign recognition , computer science , generative grammar , convolutional neural network , artificial intelligence , artificial neural network , classifier (uml) , traffic sign , sign (mathematics) , pattern recognition (psychology) , contextual image classification , machine learning , synthetic data , image (mathematics) , mathematics , mathematical analysis
In this work, we research the applicability of generative adversarial neural networks for generating training samples for a traffic sign classification task. We consider generative neural networks trained using the Wasserstein metric. As a baseline method for comparison, we take image generation based on traffic sign icons. Experimental evaluation of the classifiers based on convolutional neural networks is conducted on real data, two types of synthetic data, and a combination of real and synthetic data. The experiments show that modern generative neural networks are capable of generating realistic training samples for traffic sign classification that outperform methods for generating images with icons, but are still slightly worse than real images for classifier training.