Meta, the startup founded by Mark Zucherberg, has just published an artificial intelligence model. The model can recognise certain objects in an image.
Meta, Mark Zucherberg’s parent company, is a social media behemoth. The corporation owns popular social media platforms like as Facebook and Instagram.
In addition, the company has made an image annotation dataset public. According to the company, it is a one-of-a-kind, high-quality feature.
SAM, or Segment Anything Model, is a sophisticated artificial intelligence identification model that can find objects in both still and moving images. Therefore, even if the model hasn’t seen them during training, it can still recognise them.
How Does It Work?
Users can use the functionalities by selecting the items and clicking on them. You can also use text instructions like “dog,” “table,” or any other word.
SAM may demonstrate the needed object by drawing a square around it by accurately replying to the presented question.
According to a blog post by Meta, the model SAM is intended and taught to be promotable, so that it can transfer zero-shot to new image distributions and workloads.
“We examine its performance on a number of activities and are amazed by its zero-shot performance,” says the company. Often on par with or better than previous fully supervised results.
According to the company, the new SAM model might be used in a variety of ways by content producers. It is possible to extract images for use in video editing or collages. In addition, the model will be valuable for scientific research.
Video footage of natural events in space or on Earth can be used by researchers to locate and track animals or other objects of interest.
Furthermore, Meta already use technology to categorise photographs, remove inflammatory information, and choose which articles to recommend to Facebook and Instagram users.
According to the firm, the launch of SAM would make that technology more readily available.
A non-commercial licence will also allow you to access and download the dataset and SAM model.
To publish their images to an associated prototype, users must agree to limit their use to research purposes.
Meta has already used AI technologies that do more than just classify and identify data; they also develop totally new content.
For example, the software that makes surrealist films from word input also creates illustrations for children’s books.
“Our goal this year is to incorporate such generative AI “creative tools” into Meta’s app,” says Mark Zuckerberg.
As is commonly known, artificial intelligence has advanced society in recent months. Google and Microsoft, two technological behemoths, have also included the technology into their search engines.
Another inventive OpenAI creation is ChatGPT, AI, which has sparked a rush among large tech companies to produce analogous tools.
Nonetheless, the rapid advancement of AI has raised numerous societal and ethical concerns about a system capable of producing human-like labour as well as amazing language or images.