Abstract
Social networks have become a widely used way to share information, including images. This information can be useful for companies, which need to know their customers’ opinions and their interest in their brands. For these purposes, an image classifier has a great utility. Currently, deep learning using convolutional neural networks are heavily employed for image classification. However, they require a large number of training image samples, which are not always accessible. In order to solve this problem, we can use data augmentation, which is a regularization technique that is based on expand the original dataset to increase classification accuracy and avoid overfitting. This present work aims to compare the use of different data augmentation methods for brand logo classification. For tests, seven methods (flip, crop, rotation, gaussian filter, gaussian noise, scale, and shear) were selected based on previous studies. Two convolutional neural networks were used, AlexNet and SmallerVGGNet, this way we could see if some combinations lead to better results in different artificial network architectures. Ten different combinations of those methods show that the combination of flip, crop, rotation, and scale is a more effective combination for a brand logo classifier in the both convolutional neural networks used, improving accuracy by 22.09% using AlexNet and 12.57% using SmallerVGGNet. Other good results are found with the combination of all seven methods; flip, crop, rotation, plus gaussian noise; and the combination of flip, crop, and rotation. Further research is intended to use another regularization technique along with data augmentation to improve even more the accuracy and reduce overfitting.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsNotes
- 1.
Over-fitting of a classification model occurs when too many and/or irrelevant model terms are included and it may lead to low robustness/repeatability when the classification model is applied to independent validation data [9].
References
Gao, Y., Wang, F., Luan, H., Chua, T.S.: Brand data gathering from live social media streams. In: Proceedings of International Conference on Multimedia Retrieval, pp. 169:169–169:176. ICMR 2014. ACM, New York (2014). https://doi.org/10.1145/2578726.2578748
Goodfellow, I., Bengio, Y., Courville, A.: Deep Learning. MIT Press, Cambridge (2016). http://www.deeplearningbook.org
Han, D., Liu, Q., Fan, W.: A new image classification method using CNN transfer learning and web data augmentation. Expert Syst. Appl. 95, 43–56 (2018). https://doi.org/10.1016/j.eswa.2017.11.028, http://www.sciencedirect.com/science/article/pii/S0957417417307844
Hossain, M.S., Alhamid, M.F., Muhammad, G.: Collaborative analysis model for trending images on social networks. Futur. Gener. Comput. Syst. 86, 855–862 (2018). https://doi.org/10.1016/j.future.2017.01.030, http://www.sciencedirect.com/science/article/pii/S0167739X17301383
Hu, W., Hu, R., Xie, N., Ling, H., Maybank, S.: Image classification using multiscale information fusion based on saliency driven nonlinear diffusion filtering. IEEE Trans. Image Process. 23(4), 1513–1526 (2014). https://doi.org/10.1109/TIP.2014.2303639
Hussain, Z., Gimenez, F., Yi, D., Rubin, D.: Differential data augmentation techniques for medical imaging classification tasks (2017). https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5977656/
Ismail Fawaz, H., Forestier, G., Weber, J., Idoumghar, L., Muller, P.A.: Data augmentation using synthetic data for time series classification with deep residual networks. arXiv e-prints arXiv:1808.02455, August 2018
Molina, J.F., Gil, R., Bojacá, C., Díaz, G., Franco, H.: Color and size image dataset normalization protocol for natural image classification: a case study in tomato crop pathologies. In: Symposium of Signals, Images and Artificial Vision - 2013: STSIVA - 2013, pp. 1–5, September 2013. https://doi.org/10.1109/STSIVA.2013.6644938
Nansen, C., Geremias, L.D., Xue, Y., Huang, F., Parra, J.R.: Agricultural case studies of classification accuracy, spectral resolution, and model over-fitting. Appl. Spectrosc. 67(11), 1332–1338 (2013). http://as.osa.org/abstract.cfm?URI=as-67-11-1332
Rizvi, S.T.H., Cabodi, G., Gusmao, P., Francini, G.: Gabor filter based image representation for object classification. In: 2016 International Conference on Control, Decision and Information Technologies (CoDIT), pp. 628–632, April 2016. https://doi.org/10.1109/CoDIT.2016.7593635
Rosebrock, A.: Keras and convolutional neural networks (CNNs) (2018). https://www.pyimagesearch.com/2018/04/16/keras-and-convolutional-neural-networks-cnns/
Shijie, J., Ping, W., Peiyi, J., Siping, H.: Research on data augmentation for image classification based on convolution neural networks. In: 2017 Chinese Automation Congress (CAC), pp. 4165–4170, October 2017. https://doi.org/10.1109/CAC.2017.8243510
Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. In: International Conference on Learning Representations (2015)
Twitter: Postagem do twitter (2018). https://twitter.com/elonmusk/status/700398127044362240
Wang, F., Qi, S., Gao, G., Zhao, S., Wang, X.: Logo information recognition in large-scale social media data. Multimedia Syst. 22(1), 63–73 (2016). https://doi.org/10.1007/s00530-014-0393-x
Yan, Q., Wu, L., Zheng, L.: Social network based microblog user behavior analysis. Physica A 392(7), 1712–1723 (2013). https://doi.org/10.1016/j.physa.2012.12.008, http://www.sciencedirect.com/science/article/pii/S0378437112010540
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Moraes Machado, M., Miranda Júnior, A., de Sousa Balbino, M. (2020). A Comparative Study of Data Augmentation Methods for Brand Logo Classifiers. In: Stephanidis, C., et al. HCI International 2020 – Late Breaking Papers: Interaction, Knowledge and Social Media. HCII 2020. Lecture Notes in Computer Science(), vol 12427. Springer, Cham. https://doi.org/10.1007/978-3-030-60152-2_42
Download citation
DOI: https://doi.org/10.1007/978-3-030-60152-2_42
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-60151-5
Online ISBN: 978-3-030-60152-2
eBook Packages: Computer ScienceComputer Science (R0)