Engineering Mechanics

Proceedings Vol. 29 (2023)

ENGINEERING MECHANICS 2023

Copyright © 2023 Institute of Thermomechanics of the Czech Academy of Sciences, Prague

 

Krejsa J., Vechet S., Chen K.S.
pages 135 - 138, full text

During evaluation of convolution neural networks on the task of sign language single hand alphabet classification we have discovered that in small but not negligible number of cases the training of the network does not converge at all. This paper investigates the problem that we believe is independent of the application. While the true cause of training divergence was not discovered, we can offer the reader an easy solution from practical point of view – initialization of the first CNN layer using pretrained networks parameters.

Ownership of copyright in original research articles remains with the Authors, and provided that, when reproducing parts of the contribution, the Authors acknowledge and/or reference the Proceedings, the Authors do not need to seek permission for re-use of their material.

All papers were reviewed by members of the scientific committee.