Dataset of young deer (IRT images)

After data augmentation, the dataset contains 29,194 images with annotations in the MS COCO format for the monitored classes “animal” and “person”. The total number of objects belonging to the first category is 62,781, while 18,349 objects are annotated for the second category. In this dataset, the class “young deer” was replaced by the class “animal”, because it is not possible to reliably identify a specific wildlife species from IRT images. Nevertheless, the majority of the images predominantly contain fawns. The structure of the IRT dataset of fawns is described in detail in Table 1, which summarizes selected statistical characteristics of the annotated instances of both monitored classes. Table 1: Selected properties of objects in the training and validation dataset Property Animal class Person class Number of instances 62,781 18,349 Min. width 1 1 Max. width 240 139 Avg. width 19.03 17.60 Min. height 2 3 Max. height 321 155 Avg. height 21.07 22.09 Min. area 0.001 0.001 Max. area 18.81 4.04 Avg. area 0.22 0.15 Min. number of contour points 8 8 Max. number of contour points 400 316 Avg. number of contour points 59 69