Among the major factors that may significantly affect the power production of a photovoltaic (PV) power plant is the performance of the PV panels. Their performance depends on their maintenance and especially on keeping their surface clean to allow the panels to absorb the maximum power of the solar irradiance. To improve the maintenance process especially in remote locations, the current work focuses on an edge computing system using a camera and a Neural Network (NN) to decide the cleanness of the panels. The proposed system determines whether the panel is either clean or not. The proposed system utilizes a proprietary dataset. The execution of the inference on a Raspberry Pi Zero 2W validates the design, the edge computing performance and the accuracy.