Accurately counting unique objects in geo-tagged images presents significant challenges, particularly when the same objects appear in multiple images due to overlapping coverage areas, which can result in redundant counts. Traditional object detection and counting techniques, whether applied to images or videos, often struggle to accurately enumerate unique objects due to issues such as spatial redundancy, varying perspectives, and inconsistent object appearances across different images. This paper proposes a novel approach to address the problem of counting unique objects in geo-tagged images, specifically focusing on the challenges of distinguishing between distinct and duplicate objects. The proposed approach is characterized by two key features: 1) machine learning techniques are utilized to detect objects and improve distinguishing between them under varying imaging conditions, and 2) a spatial data structure is incorporated to organize the detected objects based on their geographical coordinates, allowing for unique object identification through spatial proximity analysis. As a case study, we apply our approach to counting homeless encampments in Los Angeles using Google Street View images, demonstrating its effectiveness in reducing counting redundancies and improving accuracy.