Face contour feature stylized generation method based on unsupervised learning

摘要: The invention provides a face contour feature stylized generation method based on unsupervised learning. comprises the following steps: acquiring aface picture, performing extraction by utilizing and key points in limiting region combining color information, binarization processing local to generate sampling image after processing; CycleGAN generative adversarial network, constructing model, importing sampled into model obtain cartoon of region. By implementing method, problems prior art are solved, stylization result robustness improved.