StEP: Style-based Encoder Pre-training for Multi-modal Image Synthesis | IEEE Conference Publication | IEEE Xplore