Green Space Quality Analysis Using Machine Learning Approaches

Green space is any green infrastructure consisting of vegetation. linked with improving mental and physical health, providing opportunities for social interactions activities, aiding the environment. The quality refers to condition space. Past machine learning-based studies have emphasized that littering, lack maintenance, dirtiness negatively impact perceived These methods assess spaces their qualities without considering human perception spaces. Domain-based methods, on other hand, are labour-intensive, time-consuming, challenging apply large-scale areas. This research proposes build, evaluate, deploy a learning methodology assessing at human-perception level using transfer pre-trained models. results indicated developed models achieved high scores across six performance metrics: accuracy, precision, recall, F1-score, Cohen’s Kappa, Average ROC-AUC. Moreover, were evaluated file size inference time ensure practical implementation usage. also implemented Grad-CAM as means evaluating heat maps. best-performing model, ResNet50, 98.98% 99.00% Kappa score 0.98, an ROC-AUC 1.00. ResNet50 model has relatively moderate was second quickest predict. visualizations show can precisely identify areas most important its learning. Finally, deployed Streamlit cloud-based platform interactive web application.