Photovoltaic (PV) power, known as solar power, operates on a simple principle known as the photovoltaic effect, in which a PV cell turns sunlight into electricity. Because PV modules are built up of series connections of PV cells, cell defects are identified as a key source of module output deterioration. Because certain defects are not obvious, even experts may fail to notice them. In this paper, we empirically analyze machine learning and deep learning models in order to offer an automatic classification technique based on electroluminescence (EL) images from PV cells. Experimental results reveal that EfficientNetB0 can be a competitive model in terms of performance and training cost.