Red Lesion Endoscopy Dataset

This dataset contains 2 subsets of anonymized video capsule endoscopy images with annotated red lesions. Folder "Set 1" has 3,295 non-sequential frames in sub-folder "A" and the corresponding annotated masks in sub-folder "B"; Similarly, folder "Set 2" has 600 sequential frames. All frames are 320x320 pixels wide. This dataset was used in the article "A Deep Learning Approach for Red Lesions Detection in Video Capsule Endoscopies" published at ICIAR 2018 to evaluate deep learning U-Net architecture, detect and segment red lesions in the small bowel.

To access the dataset please request your password via the link http://bioimglab.inesctec.pt/?page_id=2059 and fill the form.

本数据集包含两个子集，均为经过匿名化处理的视频胶囊内镜（video capsule endoscopy）图像，且附带标注的红色病变区域。文件夹"Set 1"包含子文件夹"A"中的3295张非连续帧图像，以及子文件夹"B"中对应的标注掩码文件；类似地，文件夹"Set 2"包含600张连续帧图像。所有图像的分辨率均为320×320像素。
本数据集曾被用于发表于ICIAR 2018的论文《A Deep Learning Approach for Red Lesions Detection in Video Capsule Endoscopies》，以评估深度学习U-Net架构在小肠红色病变检测与分割任务中的性能。
如需获取该数据集，请通过链接 http://bioimglab.inesctec.pt/?page_id=2059 申请密码并填写申请表单。

搜集汇总

背景与挑战

背景概述

该数据集包含两个子集的视频胶囊内窥镜图像，共计3,895帧，均标注了红色病变，用于深度学习U-Net架构的评估。图像分辨率为320x320像素，数据集由INESC TEC发布，关联于ICIAR 2018发表的研究文章。