Rooftop Solar Panel Detection from Satellite Imagery Using Detection Transformer for Indian Terrain

The paper presents a novel approach to detecting rooftop solar panels in satellite imagery, particularly within the context of India's urban and rural landscapes. The method addresses the challenges posed by densely populated urban environments, limited availability of annotated datasets, and varying image resolutions. A custom dataset was created using high-resolution images from Google Maps, annotated with labeling and further expanded using data augmentation techniques to improve model robustness. The Detection Transformer (DETR) model with a Retinanet backbone was fine-tuned for this task, achieving detection confidence levels between 99% and 100% across diverse settings, including congested urban areas and large solar farms. The results demonstrated the model's effectiveness in identifying solar panels in complex environments. Future work will focus on extending the model's capabilities to video data and deploying it in real-world applications.