SAM 3D: Behind the two-model design | AI at Meta @AIatMeta
SAM 3D: Behind the two-model design | AI at Meta  @AIatMeta
Uploaded November 2025 | Updated September 2026, 1 day ago
SAM 3D enables accurate 3D reconstruction from a single image, supporting real-world applications in editing, robotics, and interactive scene generation.

Matt, a SAM 3D researcher, explains how the two-model design makes this possible for both people and complex environments.

đź”— Read the SAM 3D Objects research paper: go.meta.me/8c08ca
đź”— Read the SAM 3D Body research paper: go.meta.me/5e60ed

--

Subscribe: youtube.com/aiatmeta?sub_confirmation=1

Learn more about our work: ai.meta.com

Follow us on Twitter: twitter.com/aiatmeta
Follow us on Facebook: facebook.com/aiatmeta
Connect with us on LinkedIn: linkedin.com/showcase/aiatmeta

Meta focuses on bringing the world together by advancing AI, powering meaningful and safe experiences, and conducting open research.
SAM 3D: Behind the two-model design | AI at MetaTeaching AI to translate 100s of spoken and written languages in real timeMSVP Improves Quality and Compression Efficiency at Metas ScaleAI for clean energy: Mike Schroepfer talks to Larry Zitnick about Open CatalystSpeaking of CICERO with Mike Lewis | Meta AIWhat does a larger vocabulary enable in Llama 3? #llama3MTIA - Metas First-Generation AI Inference Accelerator | AI at MetaAI Infra @Scale | AI at MetaInside the Lab: Building for the metaverse with AI (2022) | AI at MetaDetectron2: A PyTorch-based modular object detection libraryBuilder Bot demoThe first AI-powered speech translation system for a primarily oral language | AI at Meta
AI at Meta |

SAM 3D: Behind the two-model design | AI at Meta

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER