It was about time that Facebook’s parent firm, Meta, joined on to the artificial intelligence bandwagon. The company recently released SAM, its own language model capable of identifying different things within an image. With tools like OpenAI’s ChatGPT, Google’s Bard, and Microsoft’s new Bing, interest in AI has expanded exponentially, and Meta’s announcement reflects this. Facebook currently uses a similar system when users may tag people in postings, regulating material, and so on. Yet, Facebook’s parent firm claims that the release of SAM will “extend access to that type of technology.”
SAM (Segment Anything Model) was launched by Meta’s research division in a blog post on Wednesday.
The blog post’s title is Section Anything. “We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images,” the company noted.
The blog post goes on to say that their model is ‘built and trained to be promptable,’ that it has been examined for its capabilities, and that it has delivered ‘amazing’ results. “The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive — often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images,” the blog post read.
Not only that, but the business disclosed that SAM can recognize objects that it has not experienced during its training. A user can instruct SAM to select a certain object by either clicking on it or typing a text prompt. According to Reuters, in a demonstration of the new AI model, inputting the word “cat” caused the tool to draw boxes around each of the many cats in a photograph. Furthermore, the SAM model and dataset will be made accessible for non-commercial use. Those who upload their own photographs to an accompanying prototype must thus agree to exclusively use it for research purposes. We may learn more about Facebook’s new AI model in the coming days.