MA14KD [ORIGINAL] Dataset Description: Visual Attraction of Movie Trailers

Abstract

MA14KD [ORIGINAL] dataset ("Movie Atract 14K Dataset") provides a set of 10 VISUAL features extracted from more than 14000 movie and tv series trailers. We measured the “Attractiveness” of every frame of the movie trailers according to a paper by Jose San Pedro, and Stefan Siersdorfer and extracted the described features from movie trailers.

Full text


MA14KD [ORIGINAL] Dataset:
Visual Attraction of Movie Trailers
SUMMARY
==============================================================================
MA14KD (“Movie Atract 14K Dataset”) provides a set of 10 VISUAL features extracted from
14074 movie and tv series trailers. The movie IDs are in agreement with the movie IDs provided
by "MovieLens (ML) dataset" (ML-20M or ML Latest Version). All the movie titles, ratings and
associated movie genres and tags can be collected from the MovieLens website. We measured
the “Attractiveness” of every frame of the movie trailers according to a paper by Jose San
Pedro, and Stefan Siersdorfer

and extracted the described features from movie trailers.
1
INFORMATION ABOUT THE DATASET
==============================================================================
This dataset provides a set of 10 VISUAL features extracted from 14074 movie trailers. The data
are contained in the following comma separated file:
●MA14KD_[ORIGINAL].csv
Brief explanation of the content and the usage of these files are as follows. More details can be
found in the corresponding paper (in section CITATION). The description of each column and
each low-level visual feature is provided in the following table:
Table 1: Description of columns
Feature Type
Column title
Description
Numeric IDs
movieId
MovieLens IDs of the movie
Numeric IDs
frameId
Frame number within each movie trailer
FEATURE SET
f1
Sharpness
f2
Sharpness Variation
f3
Contrast
f4
RGB Contrast
1 https://dl.acm.org/citation.cfm?id=1526813

f5
Saturation
f6
Saturation variation
f7
Brightness
f8
Colorfulness
f9
Entropy
f10
Naturalness
DETAILS OF THE FEATURES
==============================================================================
●f1: Sharpness


measures the clarity and level of details within the elements of a frame;
●f2: Sharpness Variation is calculated via the standard deviation of all pixel sharpness
values;
●f3: Contrast measures the relative difference in brightness or color of local features in a
frame;
●f4: RGB Contrast is almost identical to the basic contrast feature, but it is extended to
the three-dimensional RGB color space;
●f5: Saturation

measures the colorfulness of the frame relative to the brightness;
●f6: Saturation variation measures the variation in saturation via the sample standard
deviation of all pixel saturation in a frame;
●f7: Brightness

measures the average brightness of a frame;
●f8: Colorfulness

measures the individual color distance of the pixels in a frame;
●f9: Entropy

measures how much information needs to be encoded by a compression
algorithm;
●f10: Naturalness measures the difference (or similarity) between a frame and the
human visual perception of the real world, with respect to colorfulness and dynamic
range.
CITATION
==============================================================================
To acknowledge the use of the dataset in publications, please cite the following paper:
Farshad B. Moghaddam, Mehdi Elahi, Reza Hosseini, Christoph Trattner, Marko Tkal
č
i
č
,
Predicting Movie Popularity and Ratings with Visual Features

, IEEE SMAP’19, 9-10 June 2019,
Larnaca, Cyprus
DOWNLOAD LINKS
==============================================================================
The MA14KD [ORIGINAL] Dataset

can be downloaded at:
●https://www.researchgate.net/publication/333579285_MA14KD_ORIGINAL_dataset

The [AGGREGATED] features of this dataset can be downloaded at:
●https://www.researchgate.net/publication/333579213_MA14KD_AGGREGATED_data
set
COLLABORATORS
==============================================================================
This database has been the result of an international collaboration. The list of collaborators is
the following:
Mehdi Elahi [Team lead]:http://www.linkedin.com/in/mehdielahi
Farshad Bakhshandegan Moghaddam: https://github.com/fmoghaddam
Reza Hosseini https://www.linkedin.com/in/reza-hoseini/
Christoph Trattner: https://www.christophtrattner.info
Marko Tkalčič:http://markotkalcic.com