top of page
Information Summary
Data information
- we downloaded a database from https://www.upf.edu/web/mtg/mard that includes 66,000 albums covers.
- we created a new structure that encompasses the relevant data - filtering out approximately 50,000 album structures that were lacking crucial information.
- we extracted the color samples from the remaining data with python imaging library and created a visual depiction of palettes with mathplotlib.
Our GitHub code: github
Problems encountered
The main problems we encountered in this project were:
Combining palettes and filtering colors-
When we compiled a palette from 120 different pictures we needed to filter colors. However, we had many different shades of colors that were difficult to combine and unite as one.
Comparison of pictures-
When comparing colors between pictures we needed to take into consideration the size of each picture and normalize it. It is incorrect to count only the 'blue' pixels in 1024X1024 picture and compare them directly to the number of pixels in 526X526 pictures.
Displaying the results-
It is impossible to represent a 2D graph that shows the total amount of color in relation to the specific color and time/per album/ per artist. (try to think about a way :) )
Calculation running time-
Comparing all matching colors between palettes requires O(N^3*P) while N is 256 and P is the number of palettes. This is not a calculation we could perform. (each color is represented by 3 numbers, 0-256. In order to compare matching colors, you need to compare each of the fields between all the palettes and limit the number of colors in the final palette. We thought we could build an array and use it as a 'bucket sort' for genre color analysis. However, we hit an out-of-memory exception, which forced us to find a way to make our solution more efficient.
​
​
Conclusions
Our main conclusions are:
​
1. White and black colors are used more often than we originally assumed.
​
2. There is a color difference between genres, and it is safe to say that different genres will usually use a slightly different color scale.
​
3. Although one might think that every artist would choose their cover according to their personal taste, the choice of color is in fact influenced by the music genre.
​
4. Bright-colored album covers are being sold for more money than the darker ones. This might suggest that the color can influence the price of the album, or consequently to point 3, that certain genres are sold for more money.
​
5. Analyzing and filtering colors is hard work.
Potential future use of this research
Future research eg. historical events based on colors
With our results, it is possible to research and view the correlation between historical events and the colors used by artists at that time.
For example, did wars cause album covers to be darker?
​
​
​
Credits and main references:
1. UPF (University Pampeu Farba)for the database
2. Cultural analytics lab- - http://lab.culturalanalytics.info/2016/04/timeline-4535-time-magazine-covers-1923.html
3. DHQ- http://www.digitalhumanities.org/dhq/vol/11/3/000332/000332.html
4. Yael Netzer for the guidance
bottom of page