Computer Vision: Interesting results from 3 different search engines
I just wanted to share a recent image search experience I had.
Problem: I saw an interesting graph in the Wall Street Journal, and I wanted to know what the name of the graph type was called. I thought, “I’ll take a screenshot of it and do an image search for it to see if I can find out what the name of this graph type is”.
Here is the screenshot of the graph:
And here are the results from the search engines.
Google thinks that it’s a brochure, I guess because of the width of the image, and the color palette. But not helpful to me.
Bing
Again, not helpful. Bing matches the color palette and the fact that it is some type of graph, but doesn’t match the specific image.
Yandex
Boom! Exactly what I was looking for. Yandex was able to match not just the color palette and the fact that it is some type of graph, but the actual graph type, ignoring color palette. Way to go Yandex! By the way, the name of this plot is an alluvial plot.
In summary, it looks like Google and Bing both matched similarity based on the content group (that it was a graph of some type) and color (light bluish) but not the actual contents of the image. Google also incorrectly tried to identify the contents of the image (that it was a brochure). Yandex nailed the correct contents of the image. However none of the search engines identified that it was an alluvial plot that I was searching for, which might be asking too much.
