Brain MRI segmentation results should always undergo a quality control (QC) process, since automatic segmentation tools can be prone to errors. In this work, we propose two deep learning-based architectures for performing QC automatically. First, we used generative adversarial networks for creating error maps that highlight the locations of segmentation errors. Subsequently, a 3D convolutional neural network was implemented to predict segmentation quality. The present pipeline was shown to achieve promising results and, in particular, high sensitivity in both tasks.