(44-2) 13 * << * >> * Russian * English * Content * All Issues

 PDF, 1210 kB

DOI: 10.18287/2412-6179-CO-601

Pages: 236-243.

Full text of article: Russian language.

Abstract:
The paper studies the possibility of using neural networks for the classification of objects that are few or absent at all in the training set. The task is illustrated by the example of classification of rare traffic signs. We consider neural networks trained using a contrastive loss function and its modifications, also we use different methods for generating synthetic samples for classification problems. As a basic method, the indexing of classes using neural network features is used. A comparison is made of classifiers trained with three different types of synthetic samples and their mixtures with real data. We propose a method of classification of rare traffic signs using a neural network discriminator of rare and frequent signs. The experimental evaluation shows that the proposed method allows rare traffic signs to be classified without significant loss of frequent sign classification quality.

Keywords:
traffic sign classification, synthetic training samples, neural networks, image recognition, image transforms, neural network compositions.

Citation:
Faizov BV, Shakhuro VI, Sanzharov VV, Konushin AS. Classification of rare traffic signs. Computer Optics 2020; 44(2): 237-244. DOI: 10.18287/2412-6179-CO-601.

1. Masana M, et al. Metric learning for novelty and anomaly detection. Source: <https://arxiv.org/abs/1808.05492>.
   
2. Zagoruyko S, Komodakis N. Wide residual networks. Source: <https://arxiv.org/abs/1605.07146>.
   
3. Krizhevsky A, Sutskever I, Hinton GE. ImageNet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems 2012; 25(2): 1097-1105.
   
4. He K, et al. Deep residual learning for image recognition. IEEE Conf Comput Soc Comput Vis Pattern Recogn 2016: 770-778.
   
5. Szegedy C, et al. Going deeper with convolutions. IEEE Conf Comput Soc Comput Vis Pattern Recogn 2015: 1-9.
   
6. Sharif Razavian A, et al. CNN features off-the-shelf: an astounding baseline for recognition. IEEE Conf Comput Soc Comput Vis Pattern Recogn Workshops 2014: 806-813.
   
7. Huang G, Larochelle H, Lacoste-Julien S. Centroid networks for few-shot clustering and unsupervised few-shot classification. Source: <https://arxiv.org/abs/1902.08605>.
   
8. Sankaranarayanan S, et al. Triplet probabilistic embedding for face verification and clustering. 2016 IEEE 8th Int Conf Biometrics: Theory Appl Syst (BTAS) 2016: 1-8.
   
9. Hadsell R., Chopra S., LeCun Y. Dimensionality reduction by learning an invariant mapping. 2006 IEEE Conf Comput Soc Comput Vis Pattern Recogn (CVPR'06) 2006; 2: 1735-1742.
   
10. Hendrycks D, Mazeika M, Dietterich TG. Deep anomaly detection with outlier exposure. Source: <https://arxiv.org/abs/1812.04606>.
    
11. HydraAPI . Source: <https://github.com/Ray-Tracing-Systems/HydraAPI>.
    
12. Shakhuro VI, Konushin AS. Image synthesis with neural networks for traffic sign classification. Computer Optics 2018; 42(1): 105-112. DOI: 10.18287/2412-6179-2018-42-1-105-112.
    
13. Zhu JY, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks. Proc IEEE Int Conf Comput Vis 2017: 2223-2232.
    
14. Huang R, et al. Beyond face rotation: Global and local perception gan for photorealistic and identity preserving frontal view synthesis. Proc IEEE Int Conf Comput Vis 2017: 2439-2448.
    
15. Shakhuro VI, Konushin AS. Russian traffic sign images dataset. Computer Optics 2016; 40(2): 294-300. DOI: 10.18287/2412-6179-2016-40-2-294-300.

---

© 2009, IPSI RAS
151, Molodogvardeiskaya str., Samara, 443001, Russia; E-mail: ko@smr.ru ; Tel: +7 (846) 242-41-24 (Executive secretary), +7 (846) 332-56-22 (Issuing editor), Fax: +7 (846) 332-56-20