This topic is apparently interesting enough that it warrants its own discussion on Quora. People there are relying on keen observational powers of human mind, but for this article, I tried to group the flags algorithmically.
I plotted the results on the map below. Countries with same colors have similar flags. The brighter the color, the bigger the group of countries with similar flags.
Here are some flag groups. To see them all, click the image above.
How I grouped the flags
I used a machine learning algorithm called k-means clustering. It’s really a rudimentary exercise, but the results are good enough to publish on this wee blog.
The algorithm accepts units to be grouped as vectors, so I had to vectorize the images first, that is to say, convert them in a long string of numbers. Each image was partitioned into a grid, then the average color value for each cell was computed. The grid was 24 x 24 cells big. I found that enough for simple flags. These color values were converted into HSB color space and experimentally weighted, then copied into a vector. These vectors were fed into the k-means algorithm with requested number of individual clusters set to 120 (there are 240 different flags). You can see results in the viewer.
Number of clusters was set experimentally, and the clustering is not perfect. For example, Canadian is grouped with some very unlikely lookalikes.
I was wondering if presence of faces in video content was an indicator of anything, and if so, of what. So I decided to scan episodes of a popular TV series and analyze them, second by second, for number of faces in video frames, and then compare charts of various episodes. Here is the result of this research.
I decided to analyze House Of Cards, partly because it’s a great series, but also because it’s character focused, so there are many scenes with a lot of people. I built an interactive viewer, which allows to see which faces were recognized at a particular point in time in Episode 3, which contains a variety of scenes with many people in them.
Launch the viewer, or continue reading for short description of technology.
Technology
To pull this off, I used the OpenCV computer vision library, which has a good capability to recognize faces. As the computer watches TV, this tool scans every frame for faces, and, if it finds any, communicates the relevant rectangles, so they can be drawn or extracted and saved.
Here’s a screenshot of a scene in church. It’s immediately apparent that the tool does not do such a good job, for many faces remain unrecognized. Still, many are recognized.
Recognized faces in the church scene, Episode 3
In this frame below, more faces are recognized.
There are also many false positives. The computer sometimes thinks that something is a face, where it most certainly it’s not, as in this picture below. If one looks carefully, one can sometimes see something face-like in these rectangles.
To construct the viewer, I extracted individual faces from frames so I could display them on the page. They are of various sizes and look like this:
To construct the charts, I just counted the faces in each seconds, then displayed the time series for each episode.
Results
This is the final chart. It’s a series of timelines that show how many faces were recognized per second. Why are some lines orange, and some yellow?
As video frames scanning progressed, some faces were recognized in only one frame in entire second – there are 23 of them. Some other faces were recognized in more frames, ans others in yet more frames. I thought this to be a good indicator of face detection reliability, but that’s not so. If it tells anything, it’s how steady the camera was in that section.
House of Cards face recognition charts by episode
My inspiration was small multiples, a visualization technique which allows for easier comparison of several datasets from the same domain. Wikipedia says:
A small multiple (sometimes called trellis chart, lattice chart, grid chart, or panel chart) is a series or grid of small similar graphics or charts, allowing them to be easily compared. The term was popularized by Edward Tufte.
According to Tufte (Envisioning Information, p. 67):
At the heart of quantitative reasoning is a single question: Compared to what? Small multiple designs, multivariate and data bountiful, answer directly by visually enforcing comparisons of changes, of the differences among objects, of the scope of alternatives. For a wide range of problems in data presentation, small multiples are the best design solution.
As always, if anyone is interested in code, mail me. My address is on About page.