The integration of solar energy into the global energy matrix is critical for mitigating climate change, reducing carbon emissions, and ensuring the sustainable management of non-renewable resources. However, precise geospatial identification of solar panel deployments remains a technical challenge, particularly in regions with diverse topographical and atmospheric conditions. Conventional satellite-based methodologies suffer from limitations in spatial resolution and susceptibility to environmental interference, hindering their effectiveness in detecting solar installations in complex landscapes such as mountainous areas, dense urban rooftops, and floating photovoltaic systems. This study proposes a deep learning-driven approach for automated solar panel detection, leveraging high-resolution imagery obtained through coordinate-based tile mapping. A carefully curated dataset of 2,200 georeferenced images—balanced between solar and non-solar classifications—was utilized for model training and evaluation. A comparative analysis of multiple convolutional neural network (CNN) architectures led to the selection of ResNet-18, optimized for its trade-off between computational efficiency and predictive accuracy. The resulting model demonstrated robustness across varied terrain and lighting conditions, achieving a validation accuracy of 98.6% and a test accuracy of 99.5%, underscoring its efficacy in large-scale solar infrastructure mapping. This study provides a scalable, automated approach to accurately mapping solar infrastructure, overcoming the limitations of traditional satellite-based detection methods in complex terrains.