Strom Impact Assessment on Banana Plantation Using Deep Learning

The project focuses on storm impact assessment on banana plantations using deep learning and image processing techniques. It leverages drone-acquired images to perform semantic segmentation to identify damaged and undamaged regions within the plantation. A pre-trained DeepLabV3 model with a ResNet-50 backbone is fine-tuned for this purpose. The segmented images are analyzed to count standing and fallen trees, estimate yield loss, and assess overall plantation health. To enhance accuracy, the approach integrates machine learning algorithms such as Cross-Entropy Loss, Adam Optimizer, and Connected Component Analysis. The system offers a fast, automated, and scalable solution for precision agriculture, enabling timely decision-making and disaster recovery planning.