Meta’s new AI model can detect any object, here’s how

Meta, formerly known as Facebook, has developed an AI model called “Segment Anything” that can identify objects in images and videos without prior training. This means the AI can detect items not included in the original data set. Users can select objects by clicking on them or using text prompts. For instance, typing “cat” will prompt the AI to highlight all the cats in a photo.

According to Reuters, this feature is a breakthrough for AI, as it eliminates the need for pre-existing data sets and enables more accurate object recognition.

How does Meta’s “Segment Anything” AI model works
The “Segment Anything” model can be combined with other models to achieve various tasks. For instance, it can assist in creating a 3D representation of an object from a single image or provide views from a mixed-reality headset. This functionality reduces the requirement for additional AI training, thereby increasing efficiency.

Meta uses comparable technology for various purposes, such as moderating prohibited content, suggesting posts, and identifying people in photos.

Although the “Segment Anything” model can handle prompts in real-time, it may experience slow performance when dealing with complex image processing tasks. According to Meta, more specialised AI tools are expected to surpass this model in their respective domains.

This AI model is unlikely to be used in robots or other devices that require quick and precise object detection. Nevertheless, models such as this can be valuable in situations where relying solely on training data is impractical.

“By sharing our research and dataset, we hope to further accelerate research into segmentation and more general image and video understanding,” Meta said in the post.

Meta will make both the AI model and dataset available for download under a non-commercial licence, meaning that they cannot be used for commercial purposes. This move is aimed at promoting research and enhancing access to technology.

FacebookTwitterLinkedin

Meta, formerly known as Facebook, has developed an AI model called “Segment Anything” that can identify objects in images and videos without prior training. This means the AI can detect items not included in the original data set. Users can select objects by clicking on them or using text prompts. For instance, typing “cat” will prompt the AI to highlight all the cats in a photo.

According to Reuters, this feature is a breakthrough for AI, as it eliminates the need for pre-existing data sets and enables more accurate object recognition.

How does Meta’s “Segment Anything” AI model works
The “Segment Anything” model can be combined with other models to achieve various tasks. For instance, it can assist in creating a 3D representation of an object from a single image or provide views from a mixed-reality headset. This functionality reduces the requirement for additional AI training, thereby increasing efficiency.

Meta uses comparable technology for various purposes, such as moderating prohibited content, suggesting posts, and identifying people in photos.

Although the “Segment Anything” model can handle prompts in real-time, it may experience slow performance when dealing with complex image processing tasks. According to Meta, more specialised AI tools are expected to surpass this model in their respective domains.

This AI model is unlikely to be used in robots or other devices that require quick and precise object detection. Nevertheless, models such as this can be valuable in situations where relying solely on training data is impractical.

“By sharing our research and dataset, we hope to further accelerate research into segmentation and more general image and video understanding,” Meta said in the post.

Meta will make both the AI model and dataset available for download under a non-commercial licence, meaning that they cannot be used for commercial purposes. This move is aimed at promoting research and enhancing access to technology.

FacebookTwitterLinkedin