Forest conservation and management are increasingly challenged by evolving societal expectations, biodiversity decline, and the impacts of climate change, requiring accurate, spatially detailed data for effective decision-making. Remote sensing and photogrammetry have become critical tools allowing detailed mapping and measurement of forests worldwide. While traditional satellite and airborne remote sensing remains important, ground-based data is becoming increasingly important for monitoring biodiversity and optimising management. Despite advances in deep learning for tree segmentation and species classification, the lack of extensive, high-quality labelled datasets is hampering development. To address this issue, TreeScanPL10K is being introduced - a dataset that surpasses previous resources in scope, comprising 10,417 individual trees, with species identified for approximately 72%. The dataset, which represents most of the forest-forming species in Central Europe, was collected in various Polish forest stands of different ages, composition and development stages. TreeScanPL10K aims to support researchers and forestry professionals by enabling the training and testing of advanced analytical tools, promoting transparency and accelerating progress in precision forestry and ecological studies.
Stereńczak et al. (Mon,) studied this question.