Kvasir-Capsule

该数据集名为Kvasir-Capsule，包含了一系列标记为胃肠道正常和异常发现的图像。这些图像种类丰富，能有效用于训练和评估卷积神经网络模型。所有图像分为正常和异常两大类，尺寸为336x336像素，且具有3个通道。值得注意的是，该数据集存在类别不平衡问题，这通过正常类别的子采样方法得到了解决。具体来说，数据集包含3,500张异常图像和42,962张正常图像。该数据集的任务是检测无线胶囊内窥镜图像中的异常。

The dataset is named Kvasir-Capsule, which contains a series of images annotated with normal and abnormal gastrointestinal tract findings. These images cover diverse categories and can be effectively utilized for training and evaluating Convolutional Neural Networks (CNNs). All images are categorized into two main classes: normal and abnormal, with each image having a resolution of 336×336 pixels and 3 color channels. Notably, this dataset suffers from a class imbalance issue, which was resolved by performing subsampling on the normal class. Specifically, the dataset consists of 3,500 abnormal images and 42,962 normal images. The task of this dataset is to detect abnormalities in wireless capsule endoscopy images.