An artificial intelligence model that can recognize specific objects in a picture has been released, according to Meta. In addition to the model, Meta has also released the largest dataset of its kind to date which includes comments on images. In a recent blog post, Meta’s research division claimed that the business has created the Segment Anything Model, a sophisticated object identification model (SAM). SAM is built to recognize items in pictures and videos even if it hasn’t seen them before during training. The paradigm enables users to choose items by clicking on them or by typing text commands like “cat.” SAM was shown to be able to precisely draw boxes around many cats in a picture.
Internally, Meta has been employing SAM-like technologies to identify pictures, filter out objectionable information, and suggest articles to Facebook and Instagram users. According to the company, the distribution of SAM will increase access to this kind of cutting-edge technology beyond their own internal operations. The SAM model and dataset are available for download from the firm under a non-commercial license. However, those who submit their own photographs to the prototype that goes with it must consent to using the tool exclusively for research purposes.
“In the future, SAM may be employed to power applications in a variety of fields that call for the identification and segmentation of every item in any image. SAM might be incorporated into bigger AI systems for a more comprehensive multimodal knowledge of the environment, such as comprehending both the visual and text content of a webpage, according to the AI research community and others.
According to Meta’s blog post, SAM might make it possible to choose an object in the AR/VR space based on the user’s gaze and then “lift” it into three dimensions. The industry titan in technology has hypothesized that SAM may have a variety of uses for content producers, including the capacity to isolate picture areas for collages or video editing. The model may also be helpful in scientific research, giving researchers to locate and track animals or objects of interest within video footage of natural occurrences on Earth or in space.