Satellite Imagery-based Prediction of Poverty using Faster R-CNN

The implementation of successful poverty reduction measures and equitable resource distribution depends on the accurate determination of poverty levels in various global locations. It is still difficult to find accurate information about global economic livelihoods, especially in underdeveloped nations. To assess poverty levels, this article suggests a unique method that makes use of deep learning and satellite imagery. It is possible to identify economic activity and use it as a measure of poverty by analyzing satellite photos taken during both day and night. This strategy provides a practical and affordable means to collect data about poverty, particularly in areas where conventional data-gathering techniques are not feasible. For this strategy to operate, a Faster Regional-Convolutional Neural Network model must be trained to understand important developmental criteria such as rooftop type, illumination sources, closeness to water sources, agricultural areas, and industrial areas. Policymakers may develop a thorough grasp of the extent of poverty in various places by fusing these factors with socioeconomic data. With a high degree of accuracy, the model determines the average of the variables under observation. For enhancing our comprehension of poverty and enabling focused measures to reduce it, this combination of deep learning and satellite images shows significant potential.