Accurate monitoring of wildlife populations is essential for effective conservation and ecological management. This study investigates the use of very high-resolution satellite imagery combined with deep learning techniques to detect African elephants across both homogeneous and heterogeneous landscapes. Leveraging convolutional neural networks (CNNs), we demonstrate automated detection of elephants with accuracy comparable to that of human observers. The approach addresses key limitations of traditional aerial surveys, including observer bias, safety concerns, and logistical challenges, while enabling repeatable, large-scale monitoring that minimizes disturbance to wildlife. Our findings highlight the growing potential of satellite-based remote sensing as a complementary tool in wildlife conservation and ecological research.