movie trailer research mathew's...
TRANSCRIPT
Temple University
School of Media and CommunicationBy Ping Feng
Project
• Is it possible to visualize digital video material?
• Why should we visualize it?
X-axis: timeline
y-axis
: brig
htnes
s
Still Image Visualization
--The luminance variation of Van Gogh’s paintings over his life time
*Note: created via ImagePlottool from softwarestudies.comled by Lev Manovach
Gone with Wind The Conversation
Digital Video Visualization
-- Cinema Redux
*Note: project developed by Brendan Dawes in 2004, picture retrieved from lab.softwarestudies.com
Color Extraction and Comparison on Movie Posts
Industry Application
Netflix Backend Database
--Color Extraction and Comparison among Different Movies
Industry Application
• Cinemetrics is thus “understood to mean the statistical analysis of quantitative data, descriptive of the structure and content of films that might be viewed as aspects of ‘style’. For practical purposes this usually means the analysis of data that describe shots in some way, for example their length or a characterization of their type. Shot lengths(SLs) have attracted most attention in the literature; examples of type includes classification by shot scale, content (e.g., action/dialogue), camera movement and so on” (Baxter, 2014, p.2).
Understanding Cinemetrics
Cinemetrics was originated from an eponymous website(www. Cinemetrics.lv) initiated by Yuri Tsivian, a professor from University of Chicago.
- The website introduces a data collection tool that works like a stopwatch that allows contributors to make annotation about the editing decision on films.
- At the same time, the website serves as a repository that displays a database of the quantified data about thousands of films (17,000, data collected till 2014) across history based on crowd collaboration. This website has attracted many scholars to do further studies by utilizing their data.
History:
Cinematrics Movie Database
Note: Polynomial trendines- retrieved from Cinemetrics.lv movie database
http://www.cinemetrics.lv/
http://cinemetrics.fredericbrodbeck.de/
http://cinemetrics.site/
Cinemetrics Dynamic Graphic Visualization
• Examine popular movie/trailer editing pattern variation and evaluation trend overtime (Project I)
• Explore media effect and audience psychological response, including attention, immersion, and presence to the editing pattern change (Project II, expected)
Research Goal and Methodology
Research Goal
Research Method
• Content analysis• Statistical analysis of quantitative data• Experiment
Basic Terminology about Video Editing
Shot Length• Shot lengths (SLs)• Average SL (ASL)• Median SL (MSL)
Shot Scale• Medium shot (MS)• Medium long shot (MLS)• Long shot (LS)• Very long shot (VLS)
Camera Movement• Big Close-up (BCU)• Close-up(CU)• Medium close-up (MC)
Frame Rate: The number of fames or images that are projected or displayed per second, usually represented as FPS, frame per second.
Shot: Between one cutting point to another cutting point in digital post-production editing decision list.
Motion: “motion is the optical change created by moving objects, people and shadows” (Cutting, et al, 2011, p.571) In this study, we use a motion indexwhich represents the percentage (value between 0.0 and 1.0) of how many pixels have significantly changed (difference above a certain threshold) from one frame to the next.
• “To establish the existence of an individual formal style in the work of director, it is necessary to compare not only a sufficient number of his film with each other, but also-which is always forgotten- to compare his films with films of similar genre made by other directors at the same time…..An even more absolute norm for any period is really needed as well, to give a standard of comparison that reflects the technical and other constraints on the work of filmmakers at that time and place…” (Salt, 1974, p.14)
• It is quantitative content analysis
• Usually examined through large data samples
What is film Style? Norm? Pattern? Does film also have pattern or norm?
Literature Review: pattern recognition, comparison, re-contextualization
Literature Review
Measured 160 English-language films released from 1935 to 2010 and found four linear changes:
• Shot lengths have gotten shorter
• More motions and movements in contemporary films than in earlier films
• Shorter shots also have proportionately more motion than longer shots, whereas there is no such relation in older films.
• Films have gotten darker; the mean luminance value of frames across the length of a film has decreased over time.
Cutting, J. E., Brunick, K. L., DeLong, J. E., Iricinschi, C., & Candan, A. (2011). Quicker, faster, darker: Changes in Hollywood film over 75 years. i-Perception, 2(6), 569.
*Notes: psychological effects associated with these four changes and suggest that all four linear trends have a single cause: Filmmakers have incrementally tried to exercise more control over the attention of filmgoers. We suggest these changes are signatures of the evolution of popular film; they do not reflect changes in film style.
Literature Review
H1: The average shot length of the movie trailers are decreasing overtime H2: The ratio of number of shot versus the shot length are increasing overtimeH3: The correlation coefficient of the shot length and motion index are significantly negativeH4: The correlation coefficient of the shot length and motion index are getting stronger overtimeH5: The luminance of the color is decreasing overtime
Research Hypothesis and Research Questions
RQ1: Does trailers’ editing pattern evolve over the past few decades?
RQ2: If the trailers editing pattern does evolve overtime, then does it align with the overall movies’ editing pattern evolution overtime – based on James Cutting’s study (2011) ?
• Action genre marked as contain a lot of motions
• Has many consecutive sequels across longer span of years
Sample Rationale
Genre Name Main Cast No. Title Age Length
Action James Bond
Actor: Sean Connery
1 Dr. No 1962 3:12
2 From Russia with Love 1963 3:47
3 Goldfinger 1964 3:08
4 Thunderball 1965 3:105 You Only Live Twice 1967 3:19
George Lazenby 6 On Her Majesty's Secret Service 1969 3:48Sean Connery 7 Diamonds Are Forever 1971 3:38
Roger Moore
8 Live and Let Die 1973 2:45
9 The Man with the Golden Gun 1974 3:08
10 The Spy Who Loved Me 1977 3:19
11 Moonraker 1979 3:43
12 For Your Eyes Only 1981 3:45
13 Octopussy 1983 3:30
14 A View to a Kill 1985 2:50
Timothy Dalton 15 The Living Daylights 1987 1:29
16 Licence to Kill 1989 1:53
Pierce Brosnan
17 GoldenEye 1995 2:51
18 Tomorrow Never Dies 1997 2:29
19 The World Is Not Enough 1999 2:18
20 Die Another Day 2002 2:25
Daniel Craig
21 Caino Royale 2006 2:23
22 Quantum of Solace 2008 2:27
23 Skyfall 2012 2:30
24 Spectre trailer #1 2015 1:15
Sampling: James Bond Movie Series Trailers
60s
70s
80s
90s
00s-10s
• YouTube- Access and download media file or any YouTube online converter tool (found by search keywords)
Data Collection
• Linux virtual machine-Automatically detect shot boundaries and output frame numbers of each shot, frame rate, motion index, color index, and other statistics
Data Visualization
Barcode Color Visualization60s James Bond
1. Dr. No (1962) 2. From Russia with Love (1963)
3. Goldfinger (1964) 4. Thunderball(1965) 5. You Only Live Twice(1967)
6. On her Majesty’s Secret Service (1969)
3.12’ 3.47’ 3.08’ 3.19’ 3.48’3.1’
*Note: graphic length= trailer length(min) x constant1.7(cm) for better visualization
Barcode Color Visualization70s James Bond
7. Diamonds Are Forever (1971)
8. Live and Let Die (1973) 9. The Man with the Golden Gun (1974)
10. The Spy Who Loved Me (1977)
11. Moonraker (1979)
3.38’ 2.45’ 3.08’ 3.19’ 3.43’
*Note: graphic length= trailer length(min) x constant1.7(cm) for better visualization
Barcode Color Visualization80s James Bond
12. For Your Eyes Only (1981)
13. Octopussy (1983) 14. A View to a Kill (1985) 15. The Living Daylights (1987)
16. License to Kill (1989)
3.45’ 3.30’ 2.50’ 1.29’ 1.53’
*Note: graphic length= trailer length(min) x constant1.7(cm) for better visualization
Barcode Color Visualization90s James Bond
17. GoldenEye (1995) 18. Tomorrow Never Dies (1997) 19. The World Is Not Enough (1999)
2.51’ 2.29’ 2.18’
*Note: graphic length= trailer length(min) x constant1.7(cm) for better visualization
Barcode Color Visualization2000s James Bond
20. Die Another Day (2002) 21. Caino Royale (2006) 22.Quantum of Solace (2008) 23.Skyefall (2012) 24.Spectre
Trailer #1
Trailer #2
2.25’ 2.23’ 2.27’ 2.23’ 1.15’
1.15’
*Note: graphic length= trailer length(min) x constant1.7(cm) for better visualization
70s: 7-11
60s:1-6Nightingale Average Color Visualization
2000s: 20-24
Trailer #1 Trailer #2
Nightingale Average Color Visualization
80s: 12-16
90s:17-19
Nightingale Average Color Visualization
Visualization explanation
Data Analysis
Number of shots/trailer length(seconds)
24 trailers over time
60s’ 70s’ 80s’ 90s’ 2000s’
0
0.5
1
1.5
2
2.5
3
3.5
-1 4 9 14 19 24
Number of Shots Versus Trailer Length Ratio
Ratio
*Note: only one trailer sample was selected for each movie, the one that is marked as official trailer available on Youtube;
• The number of shots versus trailer length ratio is gradually going up over the years until recently we see a potential of decrease, however, we need more data in future to observe the trend
Number of Shots Versus Trailer Length Ratio Change Overtime
0
0.5
1
1.5
2
2.5
3
1 3 5 7 9 11 13 15 17 19 21 23
Average Shot Length Evolution Trend Over Time
ASLLinear (ASL)
24 trailers over time
ASL (seconds)
* Note: 24th trailer #1- the slow version was removed from the data
Average Shot Length Evolution Trend Overtime
• From the graphics, we could clear see from the trend line that the average shot length is decreasing over time
* Note: 24th trailer was removed from the data
* Motion index is the percentage (value between 0.0 and 1.0) of how many pixels have significantly changed (difference above a certain threshold) from one frame to the next.
Linear regression model, P= 0.8462
Shots and Motions Correlation Coefficient Change Overtime
Correlation of Shots and Motions of Each Trailer Overtime
Correlation of Shots and Motions of Each Generation of Trailers Overtime
• The correlation coefficient of shots and motions are all negative, showing that shorter the shots are, the more motions there are
• However, the overall correlation coefficient is not significant, showing that the correlation doesn’t have too much variation over time which is not aligned with Cutting’s study
R Visualization
plot(LNY.min.cumsum$SL,LNYfit1,type="n",lwd=2,col="black",xlab="cut", ylab="SL",main="Lights of New York",ylim=c(0,max(LNY.SL.sec)))
for (i in 1:n)lines(c(minA[i],minA[i]),c(0,secA[i]),col="red")for (i in 1:n)lines(c(minD[i],minD[i]),c(0,secD[i]),col="blue")for (i in 1:n)lines(c(minT[i],minT[i]),c(0,secT[i]),col="green")
lines(LNY.min.cumsum$SL, LNYfit1, type="l",lwd=3,col="black")
# Create Wallpaperplot(LNY.min.cumsum$SL, LNYfit2,type="n",xlab="minutes",ylab="SL",main="Light of New York",ylim= (c(0,max(LNYfit1))))for(i in 1:length(minA)) abline (v=minA[i],lwd=1.5,lty=1, col="red")for(i in 1:length(minD)) abline (v= minD[i],lwd=1.5,lty=1, col="skyblue")for(i in 1:length(minT)) abline (v= minT[i],lwd=1.5,lty=1, col="green”)
lines(LNY.min.cumsum$SL,LNYfit2,lwd=4)legend ("topleft","Loess(1/9, g, 2)", lty=1, lwd=4, col="black", bty= "o", bg= "white")
Work cited:
• Baxter, M.(2014). Notes on Cinemetric Data Analysis• Baxter, M. (2013). Evolution in Hollywood editing pattern. Cinemetrics Websits.• Cutting, J.E., Brunick, K.L., BeLong, J.E., Irichischi, C., &Canadan, A.(2011), Quicker,
faster, darker: change in Hollywood film over 75 years, i-Perception, 2(6), 569
Question
Thank you