Meta’s research division released a visual object recognition model called Segment Anything Model (SAM) that is able to identify objects within images and videos using AI.
Segment Anything Model by Meta
According to a blog article published by Meta’s research division, during training, the Segment Anything Model (SAM) could recognize objects within images and videos without having seen them before.
Meta already uses SAM-like technology for internal purposes, such as to identify offensive posts, and to curate the best content for Facebook and Instagram users to see.
The firm claimed that by making SAM public, more people would have entry to such technology.
Both the SAM model and dataset will be made freely accessible for research purposes. Users who submit their own photos to a sample provided with the project must also agree to use the prototype only for research purposes.
Meta is unifying its efforts to compete with ChatGPT and OpenAI’s and in February they announced the release of LLaMA their AI language model.