A dance action standardization evaluation method and system based on pose estimation

Abstract

The application relates to a dance action standardization evaluation method and system based on pose estimation. The method comprises the following steps: extracting a two-dimensional skeleton key point sequence from a dancer performance video, and reconstructing and generating a three-dimensional pose sequence; performing time sequence alignment and spatial pose comparison on the three-dimensional pose sequence and three-dimensional data in a standard action library to generate quantized action difference data; scoring the difference data by using a network based on contrast learning to obtain an action standardization score, inputting deviation information into a strategy network based on reinforcement learning to generate personalized error correction suggestions specific to body parts, directions, amplitudes and rhythms; and comprehensively generating a standard evaluation result by combining the action standardization score and the suggestions. The method can effectively solve the problems of strong subjectivity, low efficiency and general feedback of traditional manual evaluation, and realize objective quantification, automatic high efficiency and precise guidance of dance action standardization evaluation.

Images