A new computer vision system from Microsoft can identify objects better than the average human, according to a new research paper.
Humans can identify objects with an error rate of 5.1 percent, but Microsoft's new system only has an error rate of 4.94 percent. It's a small difference, but it's never been done before.
To get these results, the research team out of Beijing had the system analyze ImageNet's 2012 classification dataset, which contains over 1.2 million training images, 50,000 validation images and 100,000 test images.
The researchers say there are still things humans are better at recognizing overall, but the system performed better in this scenario. "Humans have no trouble distinguishing between a sheep and a cow. But computers are not perfect with these simple tasks," said Jian Sun, a principal researcher at Microsoft. "However, when it comes to distinguishing between different breeds of sheep, this is where computers outperform humans. The computer can be trained to look at the detail, texture, shape and context of the image and see distinctions that can't be observed by humans."
Microsoft will be able to use this kind of technology in the future to assist with things like Bing's image search functions.