Millimeter-wave (mmWave) frequency-modulated continuous-waveform (FMCW) radar technology has become widely used for advanced driver assistance systems (ADAS) because of its ability to operate in harsh environmental conditions and provide direct measurements of range and velocity. However, the spatial resolution of an FMCW radar system is limited by the number of individual radar elements in it. While many algorithms have been developed to increase sensor array resolution for sparsely populated scenes with simplistic priors, many real-world scenes have neither the required level of sparsity nor easily described priors. In this work, we propose a system that uses deep convolutional neural networks (DCNN) to produce high-resolution radar images of realistic driving scenes. Our proposed system is able to generate radar point clouds that are five times as dense as traditional algorithms such as MUSIC and Orthogonal Matching Pursuit (OMP) from simulated radar data, enabling downstream tasks such as object detection and target classification by either a human or another neural network.