DraGan

DragGAN introduces an interactive point-based manipulation method for generative adversarial networks (GANs) that allows users to control the pose, shape, expression, and layout of generated images by dragging points to reach precise target positions. The tool uses feature-based motion supervision and a new point tracking approach to localize handle points effectively. Applicable to a wide range of categories including animals, cars, humans, and landscapes, DragGAN can generate realistic images even in challenging scenarios. This method surpasses previous approaches in image manipulation and point tracking.