Meta released a new AI tool called Segment Anything Model (SAM) that allows users to easily identify and separate different objects within images. SAM was trained on the largest image segmentation dataset containing 1.1 billion segmentation masks from 11 million licensed photos. The browser-based tool makes complex computer vision tasks more accessible to researchers. While object recognition has existed, SAM aims to be a one-stop shop for image analysis. Meta hopes this encourages further building on their model for specific use cases like biology and agriculture. The tool has limitations but could benefit Meta's virtual reality goals like detecting objects through VR headsets.
Meta released a new AI tool called Segment Anything Model (SAM) that allows users to easily identify and separate different objects within images. SAM was trained on the largest image segmentation dataset containing 1.1 billion segmentation masks from 11 million licensed photos. The browser-based tool makes complex computer vision tasks more accessible to researchers. While object recognition has existed, SAM aims to be a one-stop shop for image analysis. Meta hopes this encourages further building on their model for specific use cases like biology and agriculture. The tool has limitations but could benefit Meta's virtual reality goals like detecting objects through VR headsets.
Meta released a new AI tool called Segment Anything Model (SAM) that allows users to easily identify and separate different objects within images. SAM was trained on the largest image segmentation dataset containing 1.1 billion segmentation masks from 11 million licensed photos. The browser-based tool makes complex computer vision tasks more accessible to researchers. While object recognition has existed, SAM aims to be a one-stop shop for image analysis. Meta hopes this encourages further building on their model for specific use cases like biology and agriculture. The tool has limitations but could benefit Meta's virtual reality goals like detecting objects through VR headsets.
Meta’s New AI Tool Makes It Easier For Researchers To with the dormant realm of computer vision
Analyze Photos technologies.
“I wouldn't say that this is a new area of technology.
Object segmentation already exists so I wouldn't say this is a new capability. Fundamentally, I think their approach of using foundational models is new and the size of the dataset they're training on could be novel,” says Paul Powers, CEO and founder of Physna, a search engine for 3D objects.
But what Meta hopes is that by releasing these tools
more broadly, it’ll encourage users to build on top of their generalized model for more specific use cases in The announcement comes as the social media giant fields like biology and agriculture. increasingly diverts its attention from creating a virtual reality-based Metaverse to embed AI features across its The announcement comes simultaneously as platforms like Instagram, Facebook, Messenger and Meta reportedly plans to use generative AI for WhatsApp advertisements across Instagram and Facebook. Not wanting to miss out on the buzz around AI, in late Editing photos, analyzing surveillance footage and February, CEO Mark Zuckerberg announced that he is understanding the parts of a cell. These tasks have one creating a new product team solely focused on thing in common: you need to be able to identify and building generative AI tools like artificial personas, separate different objects within an image. Instagram filters and chat-based features in Traditionally, researchers have had to start from WhatsApp and Instagram. Zuckerberg reportedly scratch each time they want to analyze a new part of spends most of his time with the new AI team. an image. The SAM tool is built for those who don’t have the AI Meta aims to change this laborious process by being infrastructure or the data capacity to create their own the one-stop-shop for researchers and web developers models to “segment” or identify different components working on such problems. On Wednesday, the of an image, say Meta AI researchers Alexander company released an AI model, called “Segment Kirillov and Nikhila Ravi. “This is happening in real Anything Model” or “SAM” through which users can time in the browser and that makes this model much create “cutouts” or segments of any item in an image more accessible to so many more people because they by clicking on a point or drawing a box around the don't need to be able to run a lot of stuff on GPU…We object. The tool can be used in research purposes, for can enable a lot more edge use cases that some other creative editing or even to make sense of objects while methods might not allow,” Ravi says. wearing a VR headset by making it faster and more efficient to carve up different parts of an image. But there are limitations of a computer vision model trained on a database of two-dimensional images, says The tech company launched the browser-based tool to Powers. For example, for the tool to detect and select the public and also open sourced its computer vision a remote held upside down, it would need to be model, which it claims is trained on “the largest trained on different orientations of the same object. segmentation dataset” of 1.1 billion segmentation Models trained on 2D images won’t help detect masks (“masks” are different parts of an image) and 11 images that are partly covered or partially exposed, he million images licensed from a large photo company. says. This means it wouldn’t accurately identify non- Meta did not disclose which company it licensed the standardized objects through an AR/VR headset or it images from. Meta AI, the artificial intelligence wouldn’t detect partially covered objects in public research arm of the social media giant, worked with spaces if used by an autonomous vehicle 130 human annotators based in Kenya to create the manufacturer. dataset, which was made through a combination of manual and automatic labeling of a billion parts of For the company, which rebranded itself from millions of images. Facebook to Meta in late 2021 and inked its commitment to the Metaverse, the most obvious use Object recognition and computer vision technologies for this object detection tool is in its virtual reality have been around for years and are already integrated spaces such as its online VR game Horizon Worlds. in various devices such as surveillance Kirillov and Ravi say that their object detection tool cameras and drones. Amazon stores for example use can be used for “gaze-based” detection of objects object recognition to detect the items you put into through virtual reality and augmented reality your basket and autonomous vehicles use it headsets. to perceive their surroundings. Contemporary startups like Runway and incumbents like Adobe have The model can detect unknown objects and work commercialized their ability to use AI to detect and across domains with underwater, microscopic, aerial select different objects within an image for their and agricultural images. Kirillov says he was inspired creative users. As snazzy generative AI chatbots have to create a generalized image segmenting model while emerged, the goal for AI researchers at Meta was to talking to PhD researchers. “I was giving a merge the advancement in AI foundational models presentation about segmentation to some natural scientists in Berkeley and people were like ‘Okay sure, this is all cool, but I need to like count and identify trees in the photos I’ve collected for my research about fires in California,’ and so this model can do that for them,” Kirillov tells Forbes.