Oriental Female Facial Makeup Recommendation Based On Landmarks

In recent years,automatic facial makeup received more and more attention,and many facial makeup softwares emerge such as mtxiuxiu.But these softwares only allow people to try different makeups without recommending the proper makeup for them,which is inconvenient sometimes.Fortunately,makeup recommendation can provide people with suitable makeup without trying many makeups,which is time-saving and labor-saving.Now,there is not too much research on facial makeup recommendation focused on oriental female whose makeup styles are different from western female at a good extent.In this thesis,we propose a partial facial make-up recommendation method and a global facial make-up recommendation method.The two methods can meet different makeup needs of oriental female.The contributions of this work can be summarized as follows:1.Since facial make-up recommendation is a new topic,there is not public dataset,especially on oriental female.As a result,we collect oriental female face dataset and oriental female makeup dataset.2.In our partial facial make-up recommendation algorithm,we design a set of facial geometric features,together with HOG features of every facial part,to recognize the shape of every single facial part.3.In our partial facial makeup recommendation algorithm,we collect a set of oriental female facial makeup rules according to which we can get the final recommended makeup.These rules are mainly designed to makeup the defects of oriental female face.4.In our global makeup recommendation algorithm,we fuse kinds of different facial features,including HOG,VGGFACE descriptors and our geometric features.These features can integrate different facial informations to evaluate the similarity between two face images.5.We introduce human beauty estimation system and human visual perception to evaluate the efficiency of our global facial makeup recommendation and partial facial makeup recommendation.Comparing with the before-makeup facial image,the average score of after-makeup facial image using partial makeup recommendation and global makeup recommendation are 10 points and 15 points higher respectively.