Mitochondrial multi-source dataset

This dataset includes about 2 million electron microscope (TEM) images from real and generated data, containing seven sources such as animals and plants. These mitochondria are of different sizes and have different characteristics at different nanoscale. The resolution of the mitochondrial images ranges from 80 × 128 to 3730 × 4800.The name of the image folder describes a cutout to a numpy array in ZYX order in this data, the folder name prefixed with x, y, and z denotes the result of the slicing from the x, y, and z planes, respectively, and the suffixes image, mask, and output in the files under them denote the sliced image, the result of the segmentation, and the segmented mitochondrial image. The image format of the sliced image and the segmentation result is tif. The mitochondrial images are in png format.The real mitochondrial data is divided into our high-resolution mitochondrial images and low-resolution mitochondrial images from public datasets (BossDB).The synthetic data was generated via Stable Diffusion Model and contains multiple sources with a total of 500,000 images, with text labels presented in txt in various folders.This dataset will be updated in the future to promote a better understanding of mitochondrial cell biology.