Using Deep Learning Techniques to Monitor a Reforested Stand in Kenya

Effective forest management and conservation require monitoring tree attributes such as height, crown diameter, and tree count to assess growth in reforested areas. This study employs deep learning techniques to estimate these attributes using drone imagery and compares the results with ground truth data. An orthophoto generated from drone images was processed using DeepForest to detect tree crowns, enabling the extraction of crown diameters and tree counts. Tree heights were estimated using Meta's DinoV2 global canopy height model and compared to manually collected ground truth data. Evaluation metrics, including mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE), were used to assess accuracy. The findings demonstrate the potential of deep learning and remote sensing for effective forest monitoring, offering a cost-effective and scalable alternative to traditional methods. This research contributes to advancing sustainable forest management practices in reforested stands.