ED-Net
Polyp segmentation using EfficientNetB0 and Double U-Net with ASPP and SE-block attention.
ED-Net is a hybrid deep learning framework for colorectal polyp segmentation. It combines EfficientNetB0 and Double U-Net architectures with Atrous Spatial Pyramid Pooling (ASPP) and SE-block channel attention, achieving 92.26% Dice and 91.08% IoU on Kvasir-SEG with Test-Time Augmentation.
Currently under review at Biomedical Signal Processing and Control (Elsevier Q1, IF: 4.9).