Research On Beauty Recommendation Algorithm Based On Different Styles

In recent years,more and more women pay attention to automatic facial makeup technology,and there are many application systems in the beauty industry.However,these systems only provide some partial cosmetic methods and make-up test functions,which are difficult to meet the requirements of women’s makeup recommendation.At present,most of the face makeup recommendation algorithms are based on face similarity,which is not very reasonable.The makeup on the most similar face is not equal to the appropriate makeup.Therefore,this paper uses deep learning algorithm to study the female face makeup recommendation:(1)Four classic convolutional neural network models are used to supervise and train the fashion MNIST data set.After comparing and analyzing the experimental results,the main network of the local makeup recommendation algorithm and the overall makeup style recommendation algorithm based on the alexnet model and the vggnet model is selected.Then the influence of different activation functions on the training results of the alexnet model and the vggnet model is discussed and tested The most suitable activation function is selected and applied to the alexnet model and vggnet model.(2)In the work of data collection and annotation,2000 female facial images are selected and sorted to make a facial data set;based on the facial data set and the facial feature analysis API of face++ artificial intelligence platform,the facial local feature data set is collected and annotated,and the local features include face shape,eye shape and lip shape;based on certain make-up theory,this paper focuses on female facial features Based on the theoretical knowledge of makeup style,ten styles of makeup are selected,and the ten styles of makeup are transferred to the plain face images by using beauty camera software.The ten styles of makeup are obtained by using the facial value analysis API of face++ artificial intelligence platform According to the face value score,the makeup style corresponding to the highest score is used to label the plain face image to form the makeup style recommendation data set.(3)In the experiment of makeup recommendation algorithm,BP neural network and alexnet model are used to train the face local feature recognizer to classify the face shape,eye shape and lip shape,and then the local makeup recommendation data set collected in this paper is used to realize the local makeup recommendation;vggnet neural network model is used as the basic model of the overall makeup style recommendation algorithm,and the vggnet neural network model is applied to the research The BN operation and dropout layer are introduced,the full connection layer of vggnet neural network model is adjusted,and the makeup style recommendation data set is trained based on the improved vggnet model to realize the overall makeup recommendation.Face++ API and visual method are used to judge the experimental effect of local and overall makeup recommendation,the results show that the local makeup recommendation algorithm proposed in this paper improves women’s face value to a certain extent,and the makeup recommended by the overall makeup style recommendation algorithm proposed in this paper is really suitable for women.