Mangrove semantic segmentation on aerial images

In the Yucatán Peninsula, there exists a diverse array of mangroves including Rhizophora mangle, Avicennia germinans, and Laguncularia racemosa.These mangroves play a significant role in restoring naturals areas that have been damaged as a result of human activities.Furthermore, they serve as natural habitats for numerous animal and plant species.Studies have highlighted the significance of preserving and restoring these species through traditional methods.Recent advances in remote sensing and deep learning techniques have enabled the automated detection and quantification of mangroves.The application of deep neural network techniques to address computer vision challenges in the field of remote sensing is explored in this study.We focus on the detection and quantification of mangroves in remote image sensing, employing transfer learning and fine-tuning procedures with three distinct deep neural network architectures: SegNet-VGG16, U-Net, and Fully Convolutional Network (R-FCN).We applied some evaluation metrics to assess the performance of each architecture in train and test datasets, such as intersection over union (IoU), Dice coefficient, precision, sensitivity, and accuracy.In the train dataset, SegNet-VGG16 demonstrates superior precision and accuracy, while U-Net excels in IoU, Dice coefficient, and sensitivity.Conversely, R-FCN exhibits the highest sensitivity.In the test dataset, SegNet-VGG16 maintains its high precision and accuracy, while U-Net surpasses in terms of IoU and Dice coefficient, with R-FCN displaying the highest sensitivity.