This repository presents a polyp segmentation neural network implemented using the Residual U-Net architecture. The main objective of this project is to accurately segment polyps in the Kvasir-SEG dataset.
To achieve this goal, we implemented the Residual U-Net model based on the paper titled "Road Extraction by Deep Residual U-Net" (link: PDF). The Residual U-Net architecture is a modified version of the U-Net model that incorporates residual connections. These connections enable the network to effectively capture both local and global features, resulting in improved segmentation performance. The Residual U-Net architecture has proven successful in various segmentation tasks.
The Kvasir-SEG dataset, which provides annotated images of the gastrointestinal tract for polyp segmentation, is utilized in this project. The dataset is publicly available and can be obtained from the Kvasir-SEG dataset.
The trained model achieved a Dice score of 0.65 and an Intersection over Union (IoU) score of 0.51. These metrics are commonly used to evaluate the accuracy of segmentation models. A higher Dice score and IoU score indicate better segmentation performance.
Please refer to the original repository for further details and implementation code.