Artificial Intelligence creates food images from written recipes

ga(‘send’, ‘event’, ‘Uncategorized’, ‘article’, ‘article-industry-impression’, {nonInteraction: true});

Spotted: Three computer scientists at Tel Aviv University – Ori Bar El, Ori Licht, and Netanel Yosephian – have developed an Artificial Intelligence (AI) algorithm to make food images from written recipes. They fed a computer ingredients and instructions only, with no visual cues like recipe titles. The computer compared the recipe to tens of thousands of food images to generate its own image. Soups and salads came out more appetising than burgers, which had to do with the low-resolution photos used and the trouble tech has replicating clearly defined shapes. You can read their research paper here, published in January 2019.

Takeaway: Until recently, computers required short simplistic, visual descriptions to simulate realistic images. Give a computer “small white bird with short red beak” and you’ll get a bird. A little fuzzy perhaps, but a bird. Give a computer long blocks of text without visual cues and the results vary widely. The research marks an important step forward, as it shows AI’s ability to handle more abstract ideas and get creative. The research has wide-ranging implications for the publishing and media world. We recommend keeping an eye out for slightly fuzzy butternut squash recipes in your weekend magazine.

Website: www.english.tau.ac.il
Contact: www.linkedin.com/tel-aviv-university

Published in: January 2019

Source: New feed 1