Preprocessing Image Data for Deep Learning




Abstract:
Neural networks require big amount of input data in order to be properly trained, and the output and its accuracy depend on the quality of the input dataset. Most of the images used to train these networks either contain too much or not enough information, and therefore need to be preprocessed so as to reduce or even remove the noise from them, extract useful information and remove the useless ones, or apply other techniques that improve input quality for a neural network, such as super-resolution. With suitable input provided, it will be possible to create prediction models with higher precision and better accuracy. This paper gives an overview of state-of-the-art techniques for image preprocessing for different convolutional neural networks, and describes an application that demonstrates one of them.

CITATION:

IEEE format

D. Stojnev, A. Stojnev Ilić, “Preprocessing Image Data for Deep Learning,” in Sinteza 2020 - International Scientific Conference on Information Technology and Data Related Research, Belgrade, Singidunum University, Serbia, 2020, pp. 312-317. doi:10.15308/Sinteza-2020-312-317

APA format

Stojnev, D., Stojnev Ilić, A. (2020). Preprocessing Image Data for Deep Learning. Paper presented at Sinteza 2020 - International Scientific Conference on Information Technology and Data Related Research. doi:10.15308/Sinteza-2020-312-317

BibTeX format
Download

RefWorks Tagged format
Download