A synthetically generated dataset of honey samples observed by a digital microscope. The goal of this dataset is to train machine learning models to detect the pollen found in the images. This dataset contains 500 images with three different classes of particles commonly found in honey. The annotations are provided in YOLO format in a different directory. Each annotation is associated with the image by an unique ID. The images follow the naming convention "image_{ID}.png" and the annotations follow the convention "image_{ID}.txt". This dataset has been created by Sonicat Systems and is published under Creative Commons Attribution Non-Commercial Share Alike license.