Freestyle Layout-to-Image Synthesis

Layout "train bush grass railroad" "lego train bush grass railroad" "an ink painting of train bush grass railroad" "warehouse bush grass railroad" Layout "bench cat building bush furniture grass ground pavement roof sky tree window" "bench tabby cat building bush furniture grass ground pavement roof sky tree window" "a sketch of bench cat building bush furniture grass ground pavement roof sky tree window" "bench unicorn building bush furniture grass ground pavement roof sky tree window" Figure 1 . Freestyle Layout-to-Image Synthesis (FLIS) results generated by using our model. Each has two kinds of inputs: a layout of semantic masks (on the 1st column), and a text (on the top of each result). For each layout, we show three example results with edited texts (3rd-5th columns). They validate that our model is able to introduce new attributes (3rd column), styles (4th column), and objects (5th column), which are all unseen during training, in the synthesized images. The generated hornless unicorn is due to the layout constraint.