LG releases generative AI for image captioning

LG AI Research's Captioning AI is able to create description of an image. [LG CORP]

LG AI Research revealed a generative AI dubbed "Captioning AI" which can describe an image within 10 seconds in five sentences, a technology that is expected to contribute significantly to establishing an image search and management platform.

It is LG AI Research's first generative AI unveiled for commercial use.

The research arm of the LG Group said Monday it debuted Captioning AI at the Computer Vision and Pattern Recognition conference held in Vancouver, Canada.

Captioning AI is based on what’s called Zero-shot Image Captioning technology where an AI can identify traits and characteristics of an image although it has never seen it before, utilizing a massive amount of image data it has been fed with.

“The Captioning AI can improve the work efficiency and productivity of a company that has to manage vast amounts of images,” explained the company.

“It can generate an average of five sentences and 10 keywords within 10 seconds. It can process 100,000 images in two days, enabling the establishment of an image search and management system.”

LG AI Research collaborated with global stock photography service Shutterstock to gain know-how in captioning images to facilitate searches.

The Captioning AI has been offered to 10 global clients since last month and will undergo extra upgrades, according to Shutterstock CEO Sejal Amin. Amin said the system will help clients focus on creative work rather than repetitive processes.

BY JIN EUN-SOO [jin.eunsoo@joongang.co.kr]