EU-Nets: Enhanced, Explainable and Parsimonious U-Nets
This work addresses the problem of improving interpretability and stability in U-Net architectures for segmentation tasks, offering incremental advancements.
The paper tackles the limitations of traditional U-Net models by proposing EU-Nets, which enhance explainability and uncertainty estimation, achieving an average accuracy improvement of 1.389% and variance reduction of 0.83% with fewer than 0.1M parameters.
In this study, we propose MHEX+, a framework adaptable to any U-Net architecture. Built upon MHEX+, we introduce novel U-Net variants, EU-Nets, which enhance explainability and uncertainty estimation, addressing the limitations of traditional U-Net models while improving performance and stability. A key innovation is the Equivalent Convolutional Kernel, which unifies consecutive convolutional layers, boosting interpretability. For uncertainty estimation, we propose the collaboration gradient approach, measuring gradient consistency across decoder layers. Notably, EU-Nets achieve an average accuracy improvement of 1.389\% and a variance reduction of 0.83\% across all networks and datasets in our experiments, requiring fewer than 0.1M parameters.