SegCol
收藏资源简介:
SegCol数据集是由伦敦大学学院创建的,专门用于结肠镜检查中工具和折叠边缘的语义分割。该数据集包含6000张图像,来源于EndoMapper数据集,图像以40帧为一批次进行采样,标注了3种手术工具和1种结肠折叠边缘。数据集的创建过程包括手动标注和图像分类CNN筛选,旨在解决结肠镜检查中的导航和3D重建问题,特别是在视角和光照变化下的定位问题。
The SegCol dataset was developed by University College London, specifically for semantic segmentation of surgical instruments and colonic folded edges in colonoscopy procedures. It consists of 6000 images derived from the EndoMapper dataset, with images sampled in batches of 40 frames. The dataset is annotated with three categories of surgical instruments and one category of colonic folded edges. The creation of this dataset involves manual annotation and image filtering using a classification CNN, with the goal of addressing challenges in colonoscopy navigation and 3D reconstruction, particularly localization issues under varying viewing angles and illumination conditions.




