King Fahd University of Petroleum and Minerals

Saudi Data and Artificial Intelligence Authority (Sdaia)

Hambal Tella

Ali Al-Shaikhi

Shafiqur Rehman

Mohamed Mohandes

Bo Liu

A method and a system for detecting defects in solar panels. The method includes collecting an image dataset of infrared images of the at least one mounted and operational solar panel; applying the image dataset to a YOLOS-PV convolutional neural network; converting the image dataset to a set of patches; applying position embedding to each patch and generating a set of position embedded patches; applying the set of position embedded patches to a transformer block encoder; generating a set of detection tokens; applying the set of detection tokens to an array of multi-layer perceptron (MLP) heads; classifying defects in the position embedded patches; setting, by the MLP heads, a plurality of Bpred bounding boxes for each of the set of position embedded patches; and identifying a solar panel in need of repair based on the defects in the set of position embedded patches and the plurality of Bpred bounding boxes.

Growing community of inventors

Dhahran, Saudi Arabia

Hambal Tella