Dataset for Efficient and scalable training set generation for automated pollen monitoring with Hirst-type samplers

This dataset contains high-resolution pollen scan images collected from multiple geographic regions (France, Hungary, Sweden), along with expert-annotated bounding boxes in COCO format. It supports the development and evaluation of machine learning models for automated pollen detection and classification. The annotations were produced using Label Studio and are suitable for benchmarking object detection pipelines. This dataset was generated as part of the study "Efficient and scalable training set generation for automated pollen monitoring with Hirst-type samplers".