Leveraging VGG-16 and Optuna for Carbon Stock Estimation with Drone Imagery in Indonesia

Accurate carbon stock estimation is critical for environmental monitoring and management. Traditional field-based methods are constrained by scalability, making remote sensing a more viable alternative. This study employs VGG-16 and ResNet-20 for carbon stock regression using high-resolution drone imagery, optimized with Optuna for improved predictive performance. Experimental results indicate that VGG-16 achieves a higher R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score of 0.543 with lower RMSE and MAE than ResNet-20, demonstrating its superior feature extraction capability. However, dataset imbalance and high-carbon stock feature representation remain key challenges. Future research should explore hybrid models, ensemble learning, and the integration of multispectral data to enhance predictive accuracy and generalization.