FEATURES OF THE MODIFICATION OF THE INCEPTIONRESNETV2 ARCHITECTURE AND THE CREATION OF A DIAGNOSTIC SYSTEM FOR DETERMINING THE DEGREE OF DAMAGE TO RETINAL VESSELS

Computer Systems and Information Technologies, Год журнала: 2024, Номер 1, С. 27 - 32

Опубликована: Март 28, 2024

Diabetic retinopathy is a retinal disease caused by diabetes. The progression of this can lead to blindness. Every year, the number patients with increases. damage be slowed if it diagnosed early. article describes features creation neural network model and development system high accuracy rates for recognition diabetic retinopathy. advantages InceptionResNetv2 convolutional architecture are considered. This uses residual connections that help facilitate learning process. different methods reduce dimensionality feature map, making more economical in terms memory computation. has compared other networks. applied blood vessel segmentation eye images resolutions. In study, modification was carried out. use additional MaxPooling Dense layers improved speed network. Dropout layer effectively used prevent overtraining. determining degree origin implemented Python programming language. Model built using Keras library. Images from set EyePacs were processed cropping black frames Gaussian blur filter minimizing effect changing position images. During research, found 21 epochs needed achieve maximum accuracy. program calculates probability an image belonging certain class rate 1 98.6%, 2 - 98.5%, 3 98.3%, 4 98.15%, 5 98.1%.

Язык: Английский

---

Deep Learning-Based Multi-class Classification of Diabetic Retinopathy Utilizing Transfer Learning with MobileNet Architecture

Lecture notes in networks and systems, Год журнала: 2024, Номер unknown, С. 83 - 92

Опубликована: Янв. 1, 2024

Язык: Английский

---

SDRG-Net: Secure DR grading network for Real-Time decision support in IoMT environments

Biomedical Signal Processing and Control, Год журнала: 2024, Номер 100, С. 107040 - 107040

Опубликована: Окт. 7, 2024

Язык: Английский

---