Rationalization of neural network training parameters on WWII poster classification example

D. Kozhukhov; A. Vinokur; D. Arsentev; I. Arzamazov; M. Kandrashina

doi:10.1117/12.2680863

27 April 2023 Rationalization of neural network training parameters on WWII poster classification example

D. Kozhukhov, A. Vinokur, D. Arsentev, I. Arzamazov, M. Kandrashina

Proceedings Volume 12637, International Conference on Digital Transformation: Informatics, Economics, and Education (DTIEE2023); 126370G (2023) https://doi.org/10.1117/12.2680863
Event: International Conference on Digital Transformation: Informatics, Economics, and Education (DTIEE2023), 2023, Fergana, Uzbekistan

Abstract

Investigation of the possibility of studying the Bauhaus painting heritage with deep learning techniques. The efficiency of this approach is basically limited by the small size of the dataset, limited possibilities of using augmentation and significantly different nature of the images studied in comparison to typical datasets such as ImageNet. The example of classifying World War II posters using transfer learning and fine-tuning of the MobileNetV2 network illustrates the basic points regarding the use of deep learning to solve typical research problems in the Bauhaus heritage. The effects of augmentation, batch size, learning rate, and number of pre-training layers on the quality of learning are investigated. Practical recommendations useful to humanities researchers of art collections are presented.

Citation Download Citation

D. Kozhukhov, A. Vinokur, D. Arsentev, I. Arzamazov, and M. Kandrashina "Rationalization of neural network training parameters on WWII poster classification example", Proc. SPIE 12637, International Conference on Digital Transformation: Informatics, Economics, and Education (DTIEE2023), 126370G (27 April 2023); https://doi.org/10.1117/12.2680863

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available