Automatic brain MRI motion artifact detection based on end-to-end deep learning is similarly effective as traditional machine learning trained on image quality metrics

•N = 2072 brain MRI scans rated by neuroradiologists from a clinical perspective.•Lightweight 3D CNN achieves ∼94% balanced accuracy in identifying severe head motion.•SVM trained on IQMs achieves comparably high balanced accuracy of ∼88%.•No significant difference between models regarding confusion matrix, error rate, ROC.