Cali-sketch: Stroke calibration and completion for high-quality face image generation from human-like sketches

Image generation has received increasing attention because of its wide application in security and entertainment. Sketch-based face brings more fun better quality image due to supervised interaction. However, when a sketch poorly aligned with the true is given as input, existing image-to-image translation methods often cannot generate acceptable photo-realistic images. To address this problem, paper we propose Cali-Sketch, human-like-sketch photo-realistic-image method. Cali-Sketch explicitly models stroke calibration using two constituent networks: Stroke Calibration Network (SCN), which calibrates strokes facial features enriches details while preserving original intent features; an Synthesis (ISN), translates calibrated enriched sketches In way, manage decouple difficult cross-domain problem into easier steps. Extensive experiments verify that photos generated by are both faithful input sketches, compared state-of-the-art methods.