A researcher for Baidu who claimed to have beaten Google and Microsoft in an image recognition test back in May has admitted that the company cheated in the contest and apologized for its conduct.
According to Tencent, Baidu announced in mid-May that the company’s own supercomputer had delivered the best performance among research teams in the 2015 ImageNet Large Scale Visual Recognition Challenge (ILSVRC) held by the world’s leading image database designer ImageNet.
At the beginning of the month, computer scientists at ImageNet made a statement saying that Baidu had broken the rules during the test, breaching regulations governing the number of times companies were allowed to submit test results each week and allowing Baidu scientists to make greater progress.
Baidu emerged as the winner of the contest with an error rate of 4.58 percent compared to Microsoft’s 4.94 percent and Google’s 4.8 percent. After the revelations of the company’s cheating came to light, a researcher involved in the project made a statement on ImageNet’s official website apologizing for the company’s actions.
“We apologize for the mistake and are continuing to review the results. We have added a note to our research paper, Deep Image: Scaling Up Image Recognition, and will continue to provide relevant updates as we learn more,” the statement read.
Baidu is yet to officially apologize for the incident.
By Dominic Jackson