SAM represents a significant advancement in the field of computer vision, particularly in the area of image segmentation. Unlike traditional segmentation models that may require extensive training on specific object classes, SAM aims to generalize the concept of segmentation, allowing it to recognize and segment objects even in categories it has not been explicitly trained on. This capability makes SAM particularly valuable for applications requiring detailed analysis and understanding of complex visual scenes, such as autonomous driving, medical imaging, and environmental monitoring. By leveraging deep learning techniques and large-scale datasets, SAM pushes the boundaries of what's possible in object recognition and segmentation, providing more flexible and powerful tools for visual perception.
While I don't have the specific year SAM was first introduced or when it gained popularity, the trend towards "segment anything" capabilities in models has been growing, particularly in the late 2010s and early 2020s, as deep learning techniques have matured and datasets have grown.
The development of models like SAM often involves contributions from both academic researchers and industry professionals. Teams at major technology companies and leading universities are typically at the forefront of advancing computer vision technology, though without more specific information, it's difficult to pinpoint the exact individuals or groups responsible for SAM.