These data support the article 'Customized CycleGAN with Balanced Skip Networked Residual Layer (CCGAN-BSNRL) to enhance colonoscopy image by NBI translation'

The study associated with these data proposes a novel skip-connected residual layer-based customized CycleGAN model, which we refer to as Customized CycleGAN with Balanced Skip Networked Residual Layer (CCGAN-BSNRL), to achieve effective translation of WL colonoscopy images into Narrowband images. The dataset used in the study was obtained through a formal request process from Basque Biobank, and its usage is governed by a restricted agreement. According to the terms, the dataset is intended strictly for research and educational purposes, cannot be used for commercial purposes without explicit permission, and cannot be redistributed publicly. Therefore, we cannot share the dataset publicly as we are restricted from doing so. However, we share a complete Data Accessibility Statement in the paper, providing the direct link to the dataset provider's request page so other researchers may request access under the same conditions, and we include here the full code and processing pipeline for public release. The dataset contains 2131 WL and 1302 NBI images and associated polyp segmentation masks. The WL and NBI images are arranged in train, validation, and test directories, but are not paired. For translation, we subdivided each directory into subdirectories to separate WL and NBI images. Finally, in two different directories, the train directory contains 1382 WL and 821 NBI images, the validation directory contains 557 WL and 340 NBI images, and the test directory contains 192 WL and 141 NBI images. All the images are cropped to remove void portions using the void masks included in the dataset, and the final resolution of the images is set to 604*480, which is the resolution of most of the images after cropping.