Semantic Segmentation of Self-Supervised Dataset and Medical Images Using Combination of U-Net and Neural Ordinary Differential Equations

An architecture is proposed in this paper, which combines both the U-net and Neural Ordinary Differential Equations for semantic segmentation. This method consumes very lower memory and at the same time in many cases outperforms some state-of-the-art methodologies in terms of very well known performance metrics for semantic segmentation. The proposed approach is tested on three datasets, two of them are medical images and another one is self-supervised dataset. For all the datasets, the proposed approach outperforms the state-of-the-art methods with the same environmental setup.