Tree species annotations for deep learning

The dataset contains &gt;10,000 labeled tree species and the model weights for tree species detection. Format: Image tiles (.twf) with a 256 × 256 pixel size stride in the PASCAL VOC format. The labels are saved as a .xml. The RGB orthophotos are open source and can be accessed at: https://www.swisstopo.admin.ch/de/geodata/images/ortho/swissimage10.html. The methodology used to create the data is presented in Beloiu et al. 2023, Individual tree-crown detection and species identification in heterogeneous forests using aerial RGB imagery and deep learning, Remote sensing. Table 1. Tree species and the number of labels available for them. <strong>Scientific name</strong> <strong>Common name</strong> <strong>Class</strong> <strong>Total</strong> <em>Picea abies</em> Norway spruce 10 <strong>3459</strong> <em>Abies alba</em> Silver fir 11 <strong>2375</strong> <em>Pinus sylvestris</em> Scots pine 15 <strong>599</strong> <em>Fagus sylvatica</em> European beech 50 <strong>4181</strong> <strong><em>Total</em></strong> <strong><em>10614</em></strong>