#sam3d — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #sam3d, aggregated by home.social.
-
Erschreckend und faszinierend gleichzeitig.
Aber keine Angst: Ideen muss Du weiterhin selbst haben, da kann auch keine KI helfen, sondern dir diese im schlimmsten Fall stehlen.
https://youtu.be/6sYATW5gpGE?si=qFXBj_HppOTOzy1C
Aber Vorsicht: auch hier steckt wieder ein Zuckerlutscher dahinter ...
-
Erschreckend und faszinierend gleichzeitig.
Aber keine Angst: Ideen muss Du weiterhin selbst haben, da kann auch keine KI helfen, sondern dir diese im schlimmsten Fall stehlen.
https://youtu.be/6sYATW5gpGE?si=qFXBj_HppOTOzy1C
Aber Vorsicht: auch hier steckt wieder ein Zuckerlutscher dahinter ...
-
New open‑source breakthrough: SAM3D can isolate individual objects—like a tall lamp—in a point‑cloud scene, far beyond coarse class segmentation. See how promptable segmentation lets you pick out a human, a chair, or any item in a Meta‑style 3D environment. Dive into the details! #SAM3D #PromptableSegmentation #PointCloud #TallLamp
🔗 https://aidailypost.com/news/sam3d-isolates-specific-itemslike-tall-lampbeyond-broad-class
-
Meta's SAM3D model for segmenting objects from photos (images) and turning them into 3D models seems pretty clever. First, estimate a rough three-dimensional shape as a set of voxels, then add a detailed shape and texture within those voxels.
The paper is 42 pages, 20+ authors, using thousands of existing 3D models, countless hours of AI training
The things they have to do to extract the value from all that human input...
https://ai.meta.com/research/publications/sam-3d-3dfy-anything-in-images/
-
Meta's SAM3D model for segmenting objects from photos (images) and turning them into 3D models seems pretty clever. First, estimate a rough three-dimensional shape as a set of voxels, then add a detailed shape and texture within those voxels.
The paper is 42 pages, 20+ authors, using thousands of existing 3D models, countless hours of AI training
The things they have to do to extract the value from all that human input...
https://ai.meta.com/research/publications/sam-3d-3dfy-anything-in-images/