LAVT: Language-Aware Vision Transformer for Referring Image Segmentation - View it on GitHub