Google Cloud Vision API has very powerful capabilities to detect things like landmarks, labels, logos, image properties, and faces (amongst other). It is the face feature that I used for this post. It can detect the location of the faces as well as the emotional state of each. These are divided in four main ones: joy, sorrow, surprise, and anger. Each face gets a score under each one of the states that goes from very unlikely to very likely. This represents the likelihood of a face representing such emotion.
Throughout the 2,764 sites, 9,891 faces were detected. Making each site to have about 3 faces on average. What follows is a distribution of what was detected for each emotion.
Joy, without surprise, was the one that scored the highest in likelihood, with almost 60% of the pictures with very likely of representing joy. And about 10% of possible or likely. So most of the pictures that show faces on these sites are, or 70%, display joy.
Anchor Hospital www.anchorhospital.com
Sorrow, on the other hand, was almost not found in any of the pictures. 99% of the pictures were very unlikely to display it. Only 0.08% of them were possibly of sorrow.
Weathon Franciscan Healthcare www.mywheaton.org
Fulton County Health Center https://www.fultoncountyhealthcenter.org/
Surprise also scored low, which is a good thing since hospitals probably don’t want to portray this emotion to patients.
Anger is another negative emotion that was almost nonexistent, with 99% of the pictures scoring very unlikely. But there were a few faces that score either likely or very likely of anger.