The defect detection of solar cells and their modules has always been an urgent problem for the PV industry to improve the production quality and quantity.
The traditional defect detection algorithms can detect a single defect well but can not work effectively for various defects of different shapes and sizes in the actual production line, with a bottleneck in universality. Now, an object detection approach based on deep learning can automatically detect various defects on the electroluminescence (EL) pictures of solar PV modules.