Figure 2
From: DLA-Net: dual lesion attention network for classification of pneumoconiosis using chest X-ray images
Proposed DLA-Net architecture with a dual lesion attention (DLA) module based on an ImageNet-pretrained Xception Net20 architecture, consists of four main components: image processing that segments the lung field into six zones, feature extraction that uses ImageNet-pretrained Xception Net20 to extract semantic information, feature refinement that consists of a dual lesion attention (DLA) module to focus on lesion regions, and classification that classifies pneumoconiosis into one of four ILO categories using a fully connected (FC) layer. CA: channel attention, SA: spatial attention.
