On the Impact of Discrete Atomic Compression on Image Classification by Convolutional Neural Networks DOI Creative Commons
Віктор Макарічев, Vladimir Lukin,

Iryna Brysina

et al.

Computation, Journal Year: 2024, Volume and Issue: 12(9), P. 176 - 176

Published: Sept. 1, 2024

Digital images play a particular role in wide range of systems. Image processing, storing and transferring via networks require lot memory, time traffic. Also, appropriate protection is required the case confidential data. Discrete atomic compression (DAC) an approach providing image encryption simultaneously. It has two processing modes: lossless lossy. The latter one ensures higher ratio combination with inevitable quality loss that may affect decompressed analysis, particular, classification. In this paper, we explore impact distortions produced by DAC on performance several state-of-the-art classifiers based convolutional neural (CNNs). classic, block-splitting chroma subsampling modes are considered. shown each them produces quite small effect MobileNetV2, VGG16, VGG19, ResNet50, NASNetMobile NASNetLarge models. This research shows that, using approach, memory expenses can be reduced without significant degradation aforementioned CNN-based classifiers.

Language: Английский

Complexity Reduction in DAT-Based Image Processing DOI
Віктор Макарічев, Vladimir Lukin,

Iryna Brysina

et al.

Lecture notes in networks and systems, Journal Year: 2024, Volume and Issue: unknown, P. 553 - 565

Published: Jan. 1, 2024

Language: Английский

Citations

0
On the Impact of Discrete Atomic Compression on Image Classification by Convolutional Neural Networks DOI Creative Commons
Віктор Макарічев, Vladimir Lukin,

Iryna Brysina

et al.

Computation, Journal Year: 2024, Volume and Issue: 12(9), P. 176 - 176

Published: Sept. 1, 2024

Digital images play a particular role in wide range of systems. Image processing, storing and transferring via networks require lot memory, time traffic. Also, appropriate protection is required the case confidential data. Discrete atomic compression (DAC) an approach providing image encryption simultaneously. It has two processing modes: lossless lossy. The latter one ensures higher ratio combination with inevitable quality loss that may affect decompressed analysis, particular, classification. In this paper, we explore impact distortions produced by DAC on performance several state-of-the-art classifiers based convolutional neural (CNNs). classic, block-splitting chroma subsampling modes are considered. shown each them produces quite small effect MobileNetV2, VGG16, VGG19, ResNet50, NASNetMobile NASNetLarge models. This research shows that, using approach, memory expenses can be reduced without significant degradation aforementioned CNN-based classifiers.

Language: Английский

Citations

0