Meta has launched the Segment Anything Model 3 (SAM 3) and SAM 3D, enhancing its visual understanding capabilities. This release integrates advanced computer vision technologies directly into consumer platforms like Facebook Marketplace and Instagram. SAM 3 allows users to input descriptions to highlight matching objects in images, while SAM 3D reconstructs 2D photos into interactive 3D models, improving online shopping experiences. Key features like “Promptable Concept Segmentation” enable precise text-driven object identification, enhancing user engagement across Meta’s apps. The system operates efficiently, achieving high accuracy in real-time image processing. Additionally, Meta’s innovative data engine, combining AI annotators with an extensive dataset, has created one of the richest training grounds for visual segmentation. SAM 3’s applications extend beyond consumer use, aiding in wildlife research through automated video segmentation. Meta aims to democratize sophisticated visual effects for casual users while encouraging further AI advancements within the community.
Source link
