Heatmap misses the hotspot
Artificial intelligence still has a long way to go: Feng-GUI heatmap
Feng-GUI simulates human vision during the first 5 seconds of exposure to visuals, and creates heatmaps based on an algorithm that predicts what a real human would be most likely to look at. This offers designers, advertisers and creatives, a Pre-testing technology that predicts performance of an image, by analyzing levels of attention, brand effectiveness and placement, as well as breaking down the Flow of Attention.
Here's how Heatmap analyzes a photo of French actress Sophie Marceau. The software identifies a lot of heat under her armpit and totally overlooks the part that — in my unscientific view — is far more interesting:
Sat 07-Nov-09 | Permalink | Share
Comment from awgie
Time: November 12, 2009, 4:45 am
It would be interesting to learn how his program determines these "hot spots." I perused his website, and of the ten or so photos I viewed, my own points of interest differed significantly from what the heat map showed. I have found in my own experience that each individual is drawn to his or her own areas of interest when first glancing at a person or group of people. If a man is a leg man, he may not even notice what color a woman's blouse is if her legs are attractive. If a man is a breast man, he may not notice whether she is wearing a skirt or pants. For this reason, it is not possible to programmatically predict what a person will look at without knowing beforehand what that person's preference is. The same applies even when someone is looking at an image that has no people in it at all. Some people are attracted to soft curves, while others are drawn to crisp, straight lines. For some it's bright colors, and for others it's more earthy shades. Unless the program knows the viewer's likes and dislikes, it cannot accurately predict their "hot spots." Even predicting the most common "hot spots" for a group of viewers would depend on the group.