Method Of Enhancing Amatuer Dance Sequences Based On Deep Learning

Professionalism is an essential property of dancing,which is of great importance to the beauty of dances.In order to reach the desired professionalism,it requires many years of long and exhausting practice,good physical condition,musicality,but also,good understanding of the content of the choreography.During the development of 3D modeling technology,dance digitization technique,which uses motion capture devices,has played an important role in digital films and animation.However,high requirements of professional dances result in repeated process to capture motion dance sequences,which makes it prohibitive for large-scale acquisitions.If we can present a model that enhances professionalism to amateur dance movements,we can reduce the cost of capturing,which can push forward the process of researches in data-driven digital dance generation and synthesis,and promote development of related entertainment industries.The process of fixing dance sequences is named dance sequences’professional enhancement.Unfortunately,there is no method of professional enhancement.Some researches explored the metrics of professionalism of dance,which is lack of authority.Motion style transfer is similar to dance professional enhancement,which can hardly process dance sequences.Although music-driven dance synthesis is of great help of the improvement of dance professionalism using music features,it cannot maintain the limitation of original choreography.Audio alignment just think of tempo and has poor performance.The definition of dance professional enhancement is becoming a difficult and challenging task because of the lack of related method.In this paper,we present a model that enhances professionalism to amateur dance movements,which uses non-professional dance sequences as inputs,allowing the movement quality to be improved in both the space and time domains through a deep-learning network.The main contribution of this work are listed as follows.First,this work firstly classified and defined the specific metrics of dance professionalism.Second,this work is a pioneer research to enhance amateur dance sequence on fluency,rhythm and amplitude,which improved the quality of dance sequence visually.Third,we generate amateur data from professional dances taken from the AIST++dataset,which provided possibility of data augmentation.Finally,we present an novel application of music-dance synchronization.We demonstrate the effectiveness of our method by comparing it with two baseline motion transfer methods via thorough qualitative visual controls,quantitative metrics,and a perceptual study.We also provide temporal and spatial module analysis to examine the mechanisms and necessity of key components in our framework.