Your browser does not support the PubReader view.
Go to this page to see a list of supporting browsers.
The Korean Society for Journalism & Communication (KSJCS)
[ Article ]
Korean Journal of Journalism & Communication Studies - Vol. 65, No. 6, pp.452-481
ISSN: 2586-7369 (Online)
Print publication date 31 Dec 2021
Received 13 Aug 2021 Accepted 12 Nov 2021 Revised 01 Dec 2021
DOI: https://doi.org/10.20879/kjjcs.2021.65.6.011

사람에서 컴퓨터 자동화로의 연결을 위한 탐색 : 객체 인식(Object Detection) 딥러닝 알고리즘 YOLO4, 자세 인식(Pose Detection) 프레임워크 MediaPipe를 활용한 음악 프로그램의 여성 신체 대상화, 선정적 화면 검출 연구

윤호영**
**이화여자대학교 커뮤니케이션·미디어학부 조교수 hoyoungyoon@ewha.ac.kr
From Human Coding to Automated Detection : Detecting Visual Images of Female Body Objectification and Sexualized Poses from TV Music Programs Using YOLO4 and MediaPipe
Ho Young Yoon**
**Assistant Professor, Division of Communication·Media, Ewha Womans University hoyoungyoon@ewha.ac.kr

초록

본 연구는 방송 음악 프로그램을 대상으로 여성 신체 대상화 및 선정적 화면의 자동화된 검출을 위한 규칙을 찾는 연구이다. 그동안 사람에 의해 수동으로 연구되었던 화면 이미지를 자동화된 컴퓨터 알고리즘으로 추출하기 위해 어떠한 규칙을 적용해야 하는지 시험해 보았다. 연구는 2021년 3월 2째주에 지상파 3사 음악 프로그램에 방송된 걸그룹 브레이브걸스의 ‘롤린’ 댄스 화면을 대상으로 각 방송사 화면을 비교하며 진행되었다. 선행 연구에 기반하여 여성 신체가 대상화되는 이미지와 선정적 이미지를 12개의 화면 이미지로 구분하고, 방송 카메라 앵글이 이들 장면을 잡아낼 때 이전 화면과 달라지는 특질을 찾고자 하였다. 연구 방법으로는 영상 화면의 프레임 추출과 딥러닝에 기반한 YOLO4 객체 감지 알고리즘, 동작 감지 프레임워크인 MediaPipe를 활용하였다. 본 연구의 결과를 요약하면 다음과 같다. 우선, 여성 신체가 대상화되는 화면에서는 장면 전환이 일어나면서 이전 화면 대비 화면에 등장하는 사람의 수가 급격히 줄어드는 규칙이 발견되었다. 그런데, 이러한 규칙이 반드시 여성 신체가 대상화되는 화면에만 적용되는 것은 아니어서, 오히려 선정적 화면을 회피하는 방식으로도 확인되었다. 따라서, 추가적인 규칙으로 화면에 사람 얼굴이 등장여부를 동시적으로 적용해야 화면이 여성 신체를 대상화하는지 아니면 선정적 화면을 회피하는 방식의 카메라 앵글 화면인지 검출할 수 있는 것으로 나타났다. 부가적인 연구 문제로 지상파 방송 3사 화면의 대상화 및 선정적 화면 빈도를 살펴본 결과 MBC가 다른 방송사보다 상대적으로 대상화 및 선정적 화면이 적은 것으로 나타났다. 동작 검출을 해보니 MBC 화면에서 신체 하체부위의 화면이 다른 방송사보다 더 적게 검출되었다. 본 연구는 1개 걸그룹의 1개 노래에 대한 분석이라는 한계를 가지지만, 그동안 사람에 의한 수동 연구방법으로 연구된 방송 영상 이미지 연구를 자동화된 양적 연구를 통해 대용량 방송 화면에 적용하기 위해 어떠한 규칙이 적용되어야 하는지 탐색한 연구라는 의의를 가진다.

Abstract

The goal of this research is to examine patterns of objectification and sexualization of the female body in music programs on television. The study's goal is to identify rules for automated visual image detection of body objectification and sexualization. To do so, previous qualitative study findings were used to identify target images and the cutting-edge object-detection deep learning algorithm, YOLO4 (You Only Look Once), and MediaPipe, a framework for deep learning-based pose detection, were used to search for patterns. As this is a one-of-a-kind study linking body objectification and algorithm-based object detection, the case for analysis must be carefully chosen, taking into account random effects from unplanned camera movements caused by real-time broadcasting. The dance to the song 'Rollin' by the female group 'Brave Girls' had already been broadcasted earlier in 2017. Thus, the on-stage choreography and camera movements associated with the song were already known, making them suitable for research data. The study used music programs that aired on three broadcasting networks, KBS, MBC, and SBS, during the second week of March, 2021. To fine-tune the patterns, 12 screen images were selected by extracting keyframes from the song's original music videos. The study's findings are summarized below. To begin, when the scene transition is associated with a significant decrease in the number of people in the visual frame in comparison to the previous frame, it is frequently associated with female body objectification. Because body objection is associated with an emphasis on a specific body-part in the absence of a face, this is essentially a zoom-in technique transited from a wide-angle view of the scene. This rule, however, is insufficient for detecting objectified visual images; it can also be applied to screen images that avoid sexualized images in the given dance choreography. As a result, an additional rule is required to exclusively find images of female body objectification, and it is discovered that detecting human faces on the screen appears to be a good measure. In other words, unless the human faces on the screen do not appear with the scene transition that shows a dramatic decrease in the number of humans in the scene, a visual flow of images can be considered female body objectification. The study also compared the levels of objectification across broadcasting networks and found that MBC has a lower proportion of sexualized images than that of other networks. On MBC's music program, the MediaPipe framework for pose detection discovered fewer scene images with lower body parts than others. The findings of this study suggest that computer vision research can be used to detect female objectified bodies and sexual images in television programs.

Keywords:

Body Objectification, Object Detection, Pose Detection, YOLO4, MediaPipe

키워드:

MediaPipe, 객체 동작 감지, 여성 신체 대상화, 선정적 화면

Acknowledgments

This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2020S1A3A2A02095619). 이 논문은 2020년 대한민국 교육부와 한국연구재단의 지원을 받아 수행된 연구임.

References

  • Arnett, J. J. (2002). The sounds of sex: Sex in teens' music and music videos. In J. D. Brown, J. R. Steele, & K. Walsh-Childers (Eds.), Sexual teens, sexual media: Investigating media's influence on adolescent sexuality (pp. 253–264). Lawrence Erlbaum Associates Publishers.
  • Baek, G. Y., & Yoon, H. Y. (2012). Key frame analysis of network TV news covering female sexual crime victims : Vicitimization & Senstaionalism of Visual Image. Korean Journal of Journalism & Communication Studies, 65(2), 75-113. [https://doi.org/10.20879/kjjcs.2021.65.2.003]
  • Bissell, K. L., & Zhou, P. (2004). Must‐See TV or ESPN: entertainment and sports media exposure and body‐image distortion in college women. Journal of Communication, 54(1), 5-21. [https://doi.org/10.1111/j.1460-2466.2004.tb02610.x]
  • Bochkovskiy, A., Wang, C. Y., & Liao, H. Y. M. (2020). Yolov4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934.
  • Bradby, B. (1993). Sampling sexuality: gender, technology and the body in dance music. Popular music, 12(2), 155-176. [https://doi.org/10.1017/S0261143000005535]
  • Brown, J. D. (2002). Mass media influences on sexuality. Journal of sex research, 39(1), 42-45. [https://doi.org/10.1080/00224490209552118]
  • Chen, L. H., Hsu, H. W., Wang, L. Y., & Su, C. W. (2011, August). Violence detection in movies. In 2011 Eighth International Conference Computer Graphics, Imaging and Visualization (pp. 119-124). IEEE. [https://doi.org/10.1109/CGIV.2011.14]
  • Choi, Y.J., Chung Y., Yoon, H. Y., Kim, M. J., Kim, N. Y., Chen Lu, Sin, J., Lee, J., Kim, N. Y., Yeo, E., & Kang, J. (2020). Analysis of the 19th Presidential TV Debate Using Deep Learning Based Video Processing Algorithms Analysis of the frequency, facial expression and gaze. Korean Journal of Journalism and Communication Studies, 64(5), 319-372. [https://doi.org/10.20879/kjjcs.2020.64.5.009]
  • Cohen, J., & Weimann, G. (2000). Cultivation revisited: Some genres have some effects on some viewers. Communication reports, 13(2), 99-114. [https://doi.org/10.1080/08934210009367728]
  • Cyrus, C. J. (2003). Selling an image: girl groups of the 1960s. Popular Music, 22(2), 173-193. [https://doi.org/10.1017/S0261143003003118]
  • DiMaggio, P. (2015). Adapting computational text analysis to social science (and vice versa). Big Data & Society, 2(2), 2053951715602908. [https://doi.org/10.1177/2053951715602908]
  • Dalal, N., & Triggs, B. (2005, June). Histograms of oriented gradients for human detection. In 2005 IEEE computer society conference on computer vision and pattern recognition (CVPR'05) (Vol. 1, pp. 886-893). Ieee.
  • Dunbar, J. (2020). Visual media and the marketing of performers. In Women, Music, Culture (pp. 209-232). Routledge. [https://doi.org/10.4324/9780429028694-13]
  • Grabe, S., & Hyde, J. S. (2009). Body objectification, MTV, and psychological outcomes among female adolescents 1. Journal of Applied Social Psychology, 39(12), 2840-2858. [https://doi.org/10.1111/j.1559-1816.2009.00552.x]
  • Grabe, S., Ward, L. M., & Hyde, J. S. (2008). The role of the media in body image concerns among women: A meta-analysis of experimental and correlational studies. Psychological Bulletin, 134, 460–476 [https://doi.org/10.1037/0033-2909.134.3.460]
  • Ha, S., Park, B., & Lee, J. (2010). The types and characteristics of sexuality in Korean network televisions. Korean Journal of Broadcasting & Telecommunication Studies, 24(1), 73-112.
  • Joe. S. (2015). Violence of musi video: Examining gender strereo-type. The East-West Journalism, 14(7), 97-121.
  • Jones, D. C., Vigfusdottir, T. H., & Lee, Y. (2004). Body image and the appearance culture among adolescent girls and boys: An examination of friend conversations, peer criticism, appearance magazines, and the internalization of appearance ideals. Journal of Adolescent Research, 19, 323–339. [https://doi.org/10.1177/0743558403258847]
  • Kim K. H., Shim, M., Park, E (2012). A Study on Consumer Perception of Standards for Sexually Suggestive Features of Television Programs. Korean Journal of Broadcasting & Telecommunications Research, 78, 107-135.
  • Kim, H., & Yoon, Y. (2011). Study on the Sexuality in Music Programs for Adolescents aired on THree Network Broadcastigng Companies. Korean Journal of Broadcasting and Telecommunication Studies, 25(5), 49-88.
  • Kim, S. Y., & Seo, Y. S. (2011). The Relations of Exposure to Sexually Objectifying Media to Disordered Eating, Learning Flow, and Performance Anxiety Among Female High School Students. Korean Journal of Counseling and Psychotherapy, 23(4), 1137-1160.
  • Lee, S. (2011). Consumption Mode of Body Image as sign Value in Mass-culture: focused on idol groups in Korea. The Korean Journal of Cultural Sociology. 11, 193-234. [https://doi.org/10.17328/kjcs.2011.11.1.005]
  • Lee, S. (2014). Polarized Gender Display and Complexion of Sexualized Body Image in The Performances of Korean Idol-groups. Journal of Korean Women’s Studies, 30(4), 233-268.
  • Nam, J. (2021, 17 March). 17→19→21 Military ID Number ... Soldiers Boosted Bravegirls Chart. Hankyoreh Shinmun.
  • Oh, I. (2020). Computer Vision: From Basic Concepts to Mobile Application Examples. Seoul: Hanbit Academy.
  • Peng, Y. (2018). Same candidates, different faces: Uncovering media bias in visual portrayals of presidential candidates with computer vision. Journal of Communication, 68(5), 920-941. [https://doi.org/10.1093/joc/jqy041]
  • Peng, Y. (2021). What Makes Politicians’ Instagram Posts Popular? Analyzing Social Media Strategies of Candidates and Office Holders with Computer Vision. The International Journal of Press/Politics, 26(1), 143-166. [https://doi.org/10.1177/1940161220964769]
  • Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 779-788). [https://doi.org/10.1109/CVPR.2016.91]
  • Rich, M., Woods, E. R., Goodman, E., Emans, S. J., & DuRant, R. H. (1998). Aggressors or victims: Gender and race in music video violence. Pediatrics, 101(4), 669-674. [https://doi.org/10.1542/peds.101.4.669]
  • Samis, B, (2020). Back to basics: GOPs explained. Retrieved 08/11/21 from https://aws.amazon.com/blogs/media/part-1-back-to-basics-gops-explained, /
  • Sommers-Flanagan, R., Sommers-Flanagan, J., & Davis, B. (1993). What's happening on music television? A gender role content analysis. Sex roles, 28(11), 745-753. [https://doi.org/10.1007/BF00289991]
  • Strouse, J. S., & Buerkel-Rothfuss, N. L. (1987). Media exposure and the sexual attitudes and behaviors of college students. Journal of Sex Education and Therapy, 13(2), 43-51. [https://doi.org/10.1080/01614576.1987.11074908]
  • Srivastava, S., Divekar, A. V., Anilkumar, C., Naik, I., Kulkarni, V., & Pattabiraman, V. (2021). Comparative analysis of deep learning image detection algorithms. Journal of Big Data, 8(1), 1-27. [https://doi.org/10.1186/s40537-021-00434-w]
  • Voulodimos, A., Doulamis, N., Doulamis, A., & Protopapadakis, E. (2018). Deep learning for computer vision: A brief review. Computational Intelligence and Neuroscience, 2018. [https://doi.org/10.1155/2018/7068349]
  • Wallis, C. (2011). Performing gender: A content analysis of gender display in music videos. Sex Roles, 64(3-4), 160-172. [https://doi.org/10.1007/s11199-010-9814-2]
  • Yoon S. (2012). New Korean wave or post-colonial hybridity of girl group singers. Korean Journal of Broadcasting and Telecommunication Studies, 26(2), 46-86.
  • Zhou, P., Ding, Q., Luo, H., & Hou, X. (2017, June). Violent interaction detection in video based on deep learning. In Journal of physics: conference series (Vol. 844, No. 1, p. 012044). IOP Publishing. [https://doi.org/10.1088/1742-6596/844/1/012044]

Appendix

부록 I. 국내 참고 문헌

  • 김경희·심미선·박은희 (2012). 텔레비전 프로그램의 선정성 심의기준에 대한 수용자 인식 연구. <방송통신연구>, 78호, 109-135.
  • 김시연·서영석 (2011). 음악방송 노출을 통한 성적대상화 경험이 여자고등학생의 부정적 섭식행동, 수행불안 및 학습몰입에 미치는 영향. <한국심리학회지: 상담 및 심리치료>, 23권 4호, 1137-1160.
  • 김형지·윤영민 (2011). 청소년 대상 음악 프로그램의 선정성 연구: 지상파 방송 3사를 중심으로. <한국방송학보>, 25권 5호, 49-88.
  • 남지은 (2021, 3, 17). 17→19→21군번 ... 군인들이 역주행 밀어올린 브레이브걸스 ‘롤린’. <한겨레 신문>. https://www.hani.co.kr/arti/culture/culture_general/987067.html
  • 오일석 (2020). <컴퓨터 비전: 기본 개념부터 최신 모바일 응용 예까지> 서울: 한빛 미디어.
  • 윤선희 (2012). 신 한류 수용의 문화적 맥락과 여성적 정체성에 대한 문화 연구. <한국방송학보>, 26권 2호, 46-86.
  • 이수안 (2011). 대중문화에서 기호가치로서 몸 이미지의 소비양식: 아이돌 그룹을 중심으로. <문화와 사회>, 11호, 193-234.
  • 이수안 (2014). 대중문화 공연에 재현되는 양극화된 젠더 디스플레이와 몸 이미지의 성애화 방식. <한국여성학>, 30권 4호, 233-268.
  • 조수선 (2015). 뮤직비디오에 나타나는 성역할과 폭력성에 관한 연구. <동서언론>, 17호, 97-121.
  • 백지연·윤호영 (2021). 방송 뉴스가 재현하는 성범죄 피해 여성 이미지에 대한 키프레임 분석: 가상물, 자료화면을 통한 피해자다움의 재생산과 익명-실명 보도의 차이를 중심으로. <한국언론학보>, 65권 2호, 75-113.
  • 최윤정·정유진·윤호영·김민정·김나영·첸루·신주연·이주희·김나영·여은·강제원 (2020). 딥 러닝 (Deep learning) 기반 동영상 처리 알고리즘을 통한 19대 대선 TV 토론 영상분석: 후보자들의 등장빈도, 표정, 응시방향에 대한 분석. <한국언론학보>, 64권 5호, 319-372.
  • 하승태·박범길·이정교 (2010). 한국 지상파 텔레비전에 나타난 선정성의 유형 및 특성에 관한 연구. <한국방송학보>, 24권 1호, 73-112.