Deep Learning-Based Segmentation for Vegetation Coverage Analysis in Urban Green Open Space in Jakarta

Vegetation plays a vital role in urban environments by maintaining ecological balance, supporting biodiversity, and preserving environmental quality. It helps reduce air pollution, counteract the negative impacts of urbanization, and promote overall ecosystem health. However, the rapid expansion of infrastructure and human activities in Jakarta has led to a decline in green open spaces. Traditional methods for segmenting vegetation are often inefficient, expensive, and time-consuming, highlighting the need for more intuitive approaches. This study utilizes the Segment Anything Model (SAM) and DeepLabv3 deep learning-based segmentation models with a semantic segmentation technique to optimize the vegetation segmentation and assessment of vegetation coverage in urban environments of Jakarta. The evaluation results for both models show that DeepLabv3 outperforms SAM in vegetation segmentation accuracy and performance, generating more images with higher IoU scores, with the highest IoU score of 95.9% (95%). These results highlight the potential of advanced deep learning models in supporting sustainable urban planning and environmental monitoring through efficient and effective vegetation segmentation or mapping.