Artificial intelligence creates images of delicious food that doesn’t exist
A team of researchers from tel Aviv University have developed a neural network capable of reading the recipes and create the image that, in its opinion, represents the finished dish.
A team of three developers, a Bar ori e, ori Licht and Netanel Josephian, have created artificial intelligence with the help of a modified version of generative adversarial networks (GAN) called StackGAN V2 and 52 thousand combinations of images/recipes from a giant dataset recipe1M.
In fact, this artificial intelligence can take almost any list of ingredients and instructions, and find out how looks the finished dish.
However, the researchers admit that the system is far from perfect:
“The quality of the images in the dataset recipe1M lower compared with images in datasets and CUB Oxford102. This is reflected in a large number of blurred images with poor lighting conditions, “mushy pictures”, and that image does not have square shape (which complicates network training),” — said ori Bar El.
As far as we know, this is the only AI of its kind. So it is unlikely that a similar application will soon appear on your smartphone.
MDlavka — our store for your favorite readers. Technology Apple for the most pleasant prices are waiting for you every day from 10:00 to 21:00. Readers Macdigger.ru discount.