Enhancing Infrastructure Detection Using Drone Imagery: A Comparative Analysis Using Machine Learning Models with a Custom Bangladesh-Based Dataset for Improved Urban Planning

The development of cities especially in the third world countries such as Bangladesh pose some problems in the development of sustainable cities. There is also the issue of new build, both in relation to resource conservation but also in general. This work employs aerial remote sensing systems, or drone imagery more precisely, to create a rapid and inexpensive approach to assessing urban forms. Among the models used are: YOLOv5, YOLOv7, YOLOv8 (You Only Look Once), and an enhanced model for using in object detection of the infrastructure status. For the exposure and testing of the models, which delineates building, roads and other civil infrastructure features, a new dataset of Bangladesh was employed. The results confirmed that M2-d modeled improvement in detection accuracy across all the algorithms, but with an inefficiency variant. We observed that YOLOv8 and the proposed custom model demonstrated impressive performance in modeling the spatial and temporal differences of infrastructures. The models are used for temporal dynamics tracking and making adjustments to urban design predictions on temporal scales. Studying the findings of this research cannot avoid the opinion that the further addition of drone imagery to deep-learning models might be useful in urban planning in different-population cities of different development.