Meta has made both the AI model and the dataset available for download on its website with a non-commercial license.
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.
