JackHsieh/TissueMNIST-224-1pct-with-vlm-features
收藏资源简介:
该数据集是一个医学图像数据集,专注于肾脏细胞类型的分类任务。数据集中每个样本包含一个224x224像素的uint8格式图像,对应一个标签,标签涵盖8种肾脏细胞类型,包括集合管与连接小管、远端小管、肾小球内皮细胞、间质内皮细胞、白细胞、足细胞、近端小管片段和厚升支。此外,每个样本还附带一个长度为11的VLM(视觉语言模型)特征列表,用于辅助分析。数据集被分割为训练集(1,655个样本)、验证集(23,640个样本)和测试集(47,280个样本),总大小约为3.71GB,适用于机器学习模型训练和评估。
This dataset is a medical imaging dataset focused on the classification task of renal cell types. Each sample in the dataset consists of a 224×224 pixel uint8-format image paired with a corresponding label. The labels cover 8 types of renal cells, including collecting duct and connecting tubule, distal tubule, glomerular endothelial cells, interstitial endothelial cells, leukocytes, podocytes, proximal tubule fragments, and thick ascending limb. Additionally, each sample is accompanied by an 11-length Vision-Language Model (VLM) feature list for auxiliary analysis. The dataset is split into training set (1,655 samples), validation set (23,640 samples) and test set (47,280 samples), with a total size of approximately 3.71 GB, and is suitable for training and evaluating machine learning models.




