- SAM 2, the new AI model introduced by Meta, is capable of recognising and tracking any object moving in a video in real time, extending the image processing capabilities of its predecessor and opening up new opportunities for video editing and analysis.
- SAM 2’s real-time segmentation technology demonstrates AI’s ability to process moving images, accurately distinguishing between on-screen elements even when the object is moved out of the frame and then re-entered.
OUR TAKE
Meta has showed a new AI model called Segment Anything Model 2 (SAM 2), which is capable of recognising and tracking any object in a video in real time, opening up new possibilities for video editing and analysis. SAM 2’s real-time segmentation technology indicates AI’s ability to process moving images, even when objects move out of the frame and then back in again, and accurately differentiate between elements on the screen.
-Rae Li, BTW reporter
What happened
Meta has introduced an advanced AI model called SAM 2 that can recognise and track any object in a video in real time. Unlike previous SAM models that were limited to processing still images, SAM 2 extends its functionality to be able to process video content. SAM 2’s real-time segmentation technology demonstrates how AI can make great strides in processing moving images by being able to differentiate between elements on the screen even as they move around, disappear, and then re-emerge in the video. This technique has a wide range of promising applications that can benefit from it, from video editing to the development of computer vision systems, such as visual data processing in self-driving cars.
Meta shares a database of 50,000 videos in order to train the SAM 2 model. Although SAM 2 is currently open and free, this state of affairs may not be sustainable for long. Meta believes that SAM 2 has the potential to revolutionise interactive video editing and the development of computer vision systems, particularly in terms of accurately and efficiently tracking objects. Additionally, SAM 2’s real-time video segmentation capabilities provide a new perspective on the use of AI in video creation with broader implications than AI models for generating video content.
Also read: Meta’s $1.4B payout in landmark Texas privacy case
Also read: Meta releases AI Studio for enhanced social interactions
Why it’s important
Meta’s introduction of SAM 2 is significant for the video editing and analysis space as it marks a major advancement in real-time video object recognition and tracking technology. SAM 2’s real-time segmentation capabilities not only improve the efficiency and accuracy of video editing, but also open up new possibilities for the use of computer vision systems for applications such as automated driving. By being able to accurately identify and track objects in a video, SAM 2 provides a powerful tool for processing complex visual data.
The launch of SAM 2 reflects the potential of AI technology for video content creation and processing. Such technological advances will drive innovation in video content creation and have the potential to change the way we interact with video content, bringing users a richer and more personalised video experience.
