A Scrutinized Outliers Rate for One Class Classification of Green Landscape

The Green area is unarguably the most crucial part of the Sustainable Development Goals (SDG). The importance of the green regions includes preserving biodiversity and stabilizing climates. However, sophisticated mapping tool such as LIDAR is considerably expensive and also not accessible by people in common. This study offers the one-class classification of green regions using authentic drone images. The research manages to scrutinize the outliers rate based on literature. The methodology is cheaper and very applicable-the classification results in 95 % accuracy and 93 % weighted-average F1 score. The technology behind the method includes a lightweight neural-network architecture, a weighted Huber loss and a final Softmax function. The results show that this study is promising for future use.