The images within this repository contains raw, unprocessed, microscopic images of soil fungi belonging to the following genera: Fusarium, Trichoderma, Verticillium, and Chromista of the Phytophthora genus that filenames contain name of the genus followed by underscore and number of an image for a given genera. Each image is labeled based on the specific microorganism genus it represents. The subimages images contains retrieved subimages, where each subimage corresponds to a single object that contains fragments of a microorganism. The subimages are generated from the original images and follow a specific naming convention. The image name indicates the microorganism genus, followed by information about the input image it was generated from, and finally the number of the retrieved subimage from that input image. These subimages are utilized for training purposes using the Transfer Learning method with the CNN model called ResNet50. This repository is provided as a Supplementary Material to the article "Automated Identification of Soil Fungi and Chromista through Convolutional Neural Networks" submitted to the Engineering Applications of Artificial Intelligence journal.