Learning to reconstruct botanical trees from single images

We introduce a novel method for reconstructing the 3D geometry of botanical trees from single photographs. Faithfully tree single-view sensor data is challenging and open problem because many possible exist that fit tree's shape observed view. address this challenge by defining reconstruction pipeline based on three neural networks. The networks simultaneously mask out in input photographs, identify species, obtain its radial bounding volume - our representation trees. Radial volumes (RBV) are used to orchestrate procedural model primed learned parameters grow matches main branching structure overall captured tree. While RBV allows us faithfully reconstruct structure, we use model's morphological constraints generate realistic crown. This number solutions models given photograph show reconstructs various species even when front complex backgrounds. Moreover, although have been trained synthetic with augmentation, performs well real evaluate reconstructed geometries several metrics, including leaf area index maximum distances.