FinnWoodlands Dataset

While the availability of large and diverse datasets has contributed to significant breakthroughs in autonomous driving indoor applications, forestry applications are still lagging behind new forest would most certainly contribute achieving progress development data-driven methods for forest-like scenarios. This paper introduces a dataset called \textit{FinnWoodlands}, which consists RGB stereo images, point clouds, sparse depth maps, as well ground truth manual annotations semantic, instance, panoptic segmentation. \textit{FinnWoodlands} comprises total 4226 objects manually annotated, out 2562 (60.6\%) correspond tree trunks classified into three different instance categories, namely "Spruce Tree", "Birch "Pine Tree". Besides trunks, we also annotated "Obstacles" instances semantic stuff classes "Lake", "Ground", "Track". Our can be used where holistic representation environment is relevant. We provide an initial benchmark using models segmentation, completion, illustrate challenges that such unstructured scenarios introduce.