AI detects objects in images by using computer vision techniques that analyze the visual features of an image. The process typically involves using a convolutional neural network (CNN) to identify patterns in the image data that correspond to different object categories. The CNN first extracts features from the image using a series of convolutional and pooling layers and then uses these features to make predictions about the presence of different objects. The network is typically trained on large datasets of labelled images to learn how to recognize objects accurately in a wide variety of contexts, and can achieve high levels of accuracy in detecting objects in new images. However, Meta is moving out of the conventional method of AI detection, as its new AI, Segment Anything, can detect never-seen-before objects in images.
As per a research paper published by the social media giant, this model can be used by selecting objects by clicking them or using free-form text prompts. By simply typing “bird,” the AI will highlight all the birds in a given photo. Furthermore, the model can work in close collaboration with other models to help reconstruct an object in 3D or draw from views from a mixed-reality headset.
While the existing model may have limitations, such as missing finer details and not being as accurate at detecting boundaries as other models, it can help in situations where relying solely on training data is impractical. For instance, a social network could use the tech to keep up with a rapidly growing volume of content. Though this AI may not be utilized in devices requiring fast and accurate object detection, models like this one demonstrate the potential to limit the need for additional AI training. Meta has made both the AI model and dataset available for download with a non-commercial license. This move will expand access to the technology, primarily for research purposes. Nonetheless, this breakthrough shows that Meta wants to generalize computer vision, and it is already planning generative AI “personas” for its social apps, demonstrating that it has a few advantages of its own.
RELATED:
- Meta Rolls Out Paid Verification for Instagram & Facebook Users in the US
- Meta is Ending NFT Support for Facebook & Instagram
- Best 4K Streaming Devices
(Source)
