An empirical study on the automatic recognition and measurement of Chinese film shot type based on Mediapipe

This research integrates classic theories from film studies with cutting-edge technologies from computer science, aiming to conduct an in-depth exploration of the Chinese film. It focuses on the interdisciplinary field of shot type recognition, exploring the synergy between humanities and social science research and digital technology innovation. Initially, we discussed the standards for shot type determination and the influencing factors. Subsequently, we conducted a comparative study of two algorithms for shot type determination: one based on facial data and the other based on pose data. The research finds that shot type determination based on facial data is questionable due to factors such as age, shooting angle, character posture, and recognition rate. In contrast, although the method based on pose data has drawbacks such as misidentification, over-identification, and non-identification, it is currently the best applicable method with fast computation speed, high recognition rate, and robustness. To validate and implement these algorithms, we utilized computational resources provided by the Ningbo Artificial Intelligence Supercomputing Center to process and statistically analyze a large dataset of Chinese mainland films from 1949 to 2019. After calculating the shot types for 1,585 film samples, the study reveals that the proportion of Close Up Shot and Extreme Close Up Shot types in films has shown an overall increasing trend, while the proportion of Full Shot and Medium Shot types has shown a decreasing trend. The proportion of Long Shot types has remained relatively stable, consistently around 15%. This research advances the digital transformation of film studies by providing robust technical support and a theoretical framework for film content analysis and film and television production optimization. Furthermore, it equips researchers in computational film studies with novel tools and methods that enhance efficiency and objectivity, ultimately expanding the application of computer science across the arts and culture domain.