The rapid growth of solar photovoltaic (PV) systems as green energy sources has gained momentum in recent years. However, the anomalies of PV panel defects can reduce its efficiency and minimize energy harvesting from the plant. The manual inspection of PV panel defects throughout the plant is costly and time-consuming. Thus, implementing more intelligent ways to inspect solar panel defects will provide more benefits than traditional ones. This study presents an implementation of a deep learning model to detect solar panel defects using an advanced object detection algorithm called You Look Only Once, version 7 (YOLOv7). YOLO is a popular algorithm in computer vision for classification and localization. The dataset utilized in this study was sourced from ROBOFLOW, consisting of 1660 infrared images showcasing thermal defects in PV panels. The model was constructed to identify a broader range of images with heterogeneity, leveraging the aforementioned dataset. Following validation, the model demonstrates a mean Average Precision (mAP) of 85.9%. With this accuracy, the model is relevant for real-world applications. This assertion is affirmed by testing the model with additional data from separate video-capturing PV panels. The video was recorded using a drone equipped with a thermal camera.