Decoder architecture based on the UNet++. Combining residual bottlenecks with depthwise convolutions and attention mechanisms, it outperforms the UNet++ in a coronary artery segmentation task, while being significantly more computationally efficient.
deep-learning
pytorch
segmentation
unet
medical-image-segmentation
efficientnet
unetplusplus
efficientunetplusplus
-
Updated
Jan 20, 2022 - Python