Drone Assisted Forest Structural Classification of Kejimkujik National Park using Deep Learning

The wide array of terrestrial forest and wooded lands is one of the richest sources of biodiversity because of inherent structural diversity. Structure plays a significant role in a diversity indicator of a forest. We propose a transfer learning framework consisting of the ResNet-50 architecture for which we obtained a test accuracy of 75.86%. Analysis of structural diversity of the Kejimkujik National Park was done with the help of a drone using deep learning methods that predict the structural class of the forest. We used a novel forest structural diversity dataset collected using DJI Mavic drone to train the deep learning model.