Research on Deep Learning-based Image Processing and Classification Techniques for Complex Networks

Applied Mathematics and Nonlinear Sciences, Год журнала: 2025, Номер 10(1)

Опубликована: Янв. 1, 2025

Abstract Image classification task is a fundamental problem in the field of computer vision. With rapid development Internet and artificial intelligence technology, large amount image data generated every day. In this paper, for invalid feature information process semantic segmentation images, loss local detail paper proposes an encoder based on DCNN, ECANet DSA_ASPP. Based above encoder, algorithm lightweight multi-scale attention fusion proposed.After analyzing comparing commonly used extraction algorithms, SIFT features are as nodes network similarity measures analyzed, correlation coefficients weights connected edges network.The average intersection concurrency ratios reach 69.6% 73.6%, respectively. Compared to existing state-of-the-art models, detection performance paper’s method better, which can effectively capture reduce pixel errors. Finally, PreactResNet two benchmark datasets, CUB-200-2011 Stanford Dogs, outperforms performance.

Язык: Английский

Diagnosis of pneumonia from chest X-ray images using YOLO deep learning

Frontiers in Neurorobotics, Год журнала: 2025, Номер 19

Опубликована: Апрель 28, 2025

Early and accurate diagnosis of pneumonia is crucial to improve cure rates reduce mortality. Traditional chest X-ray analysis relies on physician experience, which can lead subjectivity misdiagnosis. To address this, we propose a novel method using the Fast-YOLO deep learning network that introduced. First, constructed dataset containing five categories applied image enhancement techniques increase data diversity model’s generalization ability. Next, YOLOv11 structure was redesigned accommodate complex features images. By integrating C3k2 module, DCNv2, DynamicConv, effectively enhanced feature representation reduced computational complexity (FPS increased from 53 120). Experimental results subsequently show our outperforms other commonly used detection models in terms accuracy, recall, mAP, offering better real-time capability clinical application potential.

Язык: Английский

MNet-SAt: A Multiscale Network with Spatial-enhanced Attention for segmentation of polyps in colonoscopy

Biomedical Signal Processing and Control, Год журнала: 2024, Номер 102, С. 107363 - 107363

Опубликована: Дек. 24, 2024

Язык: Английский