ISIT-UMR Multimodal Polyp Classification Dataset

该数据集包含了76个息肉样本，每个样本都有白光和窄带成像（NBI）两种模式下的视频，这些视频从同一息肉的不同角度捕捉纹理和颜色信息。数据集的标注基于组织病理学结果以及多位胃肠病专家的确认。此外，该数据集涵盖了3种息肉类别：15个锯齿状腺瘤、21个增生性病变和40个腺瘤。每个视频大约30秒长，以每秒26帧的速度录制。数据规模方面，共有76个息肉样本，其中53个用于训练，23个用于测试。该任务的目的是进行内镜数据中的息肉分类。

This dataset includes 76 polyp samples, each accompanied by videos captured under two imaging modalities: white light and narrow-band imaging (NBI). These videos acquire texture and color information of the same polyp from multiple perspectives. The annotations of the dataset are based on histopathological findings and confirmed by multiple gastroenterologists. Additionally, the dataset covers three polyp categories: 15 serrated adenomas, 21 hyperplastic lesions, and 40 adenomas. Each video has a duration of approximately 30 seconds and was recorded at 26 frames per second. In total, there are 76 polyp samples, with 53 allocated for training and 23 for testing. The goal of this task is polyp classification from endoscopic data.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于息肉分类的多模态内镜数据集，包含76个息肉样本，每个样本提供白光和窄带成像（NBI）两种模式的视频，以捕捉纹理和颜色信息。数据集基于组织病理学标注，涵盖锯齿状腺瘤、增生性病变和腺瘤三类，数据划分为53个训练样本和23个测试样本，旨在支持内镜息肉分类任务。