Meta has introduced a new AI model that can detect objects in images and videos, even if it wasn't trained to recognise them.
The "Segment Anything" AI model can not only recognise objects, it can also let users select each individual object by clicking on them or by using text prompts.
Also Read | WhatsApp is testing a new text editor in beta
As Reuters reports, If a user uploads an image, and types "cat", then the AI will recognise all the animals that match the description in the given photo.
It can also be paired with other AI models, and can help with 3D reconstruction of objects from an image, or even use an AR or VR display to draw from.
Meta has released both the AI model and dataset under a "non-commercial licence", which means they can be used for free, but not in commercial products meant for sale.
The model is primarily meant for research, and improving the technology. Meta says it uses similar technology to moderate content, recommend posts and recognise objects in tagged photos.
Also Read | Bipartisan US lawmakers introduce bill aimed at Google, Facebook ad clout
The social media company admits that the AI still needs work, and will sometimes miss finer details, or won't be as accurate detecting objects in some cases.
Meta also says that while the model can recognise text prompts in real-time, it will get bogged down and become slow to respond, if there are too many prompts in queue.
Discover the latest Business News, Sensex, and Nifty updates. Obtain Personal Finance insights, tax queries, and expert opinions on Moneycontrol or download the Moneycontrol App to stay updated!
