Apple Publishes Its First Paper On Artificial Intelligence
It has been not even a month when Apple said that its going to publish research papers on Artificial Intelligence. Keeping up its promise, Apple has published its research paper on December 22 which was submitted for publication on November 15.
The paper describes techniques which can improve image recognition. It suggests using computer-generated images rather than real-world images.
Paper explains this by saying data of the synthetic image are already labelled and annotated, but in the case of real-world image somebody has to do a full-scale labelling for the computer to see. The paper from Apple says Thesynthetic imagedata is often not realistic enough, leading the network to learn details only present in synthetic images and fail to generalize well on real images,”
The paper suggests something called Simulated + Unsupervised learning, which will help in synthetic image data to improve training. Apple researchers use a new machine learning technique called as Generative Adversarial Network which they use to generate the more realistic image.
The papers lead author is Ashish Shrivastava, a PhD holder in computer vision. And co-authors are Tomas Pfister, Oncel Tuzel, Wenda Wang, Russ Webb and Josh Susskind. These co-authors of the paper, co-found an AI startup called the Emotient which evaluate a persons emotions by looking at their facial expression. This AI startup was acquired by Apple last year.