WorldFoundry
English中文

VGGT-Omega

Method overview: Architecture Overview. VGGT- Ω appends camera and scene tokens (registers) to image tokens, and then alternates between global attention (or register attention) and frame attention layers. We replace the redundant.

VGGT-Omega from Meta AI: 3d-reconstruction and gaussian-splatting; Verified.

Edit this pageGitHubMarkdown

VGGT-Omega is the extended VGGT release published under facebook/VGGT-Omega, covering 3D reconstruction, Gaussian splatting, and novel view synthesis. In this catalog it is verified with unusually strong evidence: single-image GPU parity against the pinned upstream commit passed with max_abs_diff=0.0 for comparable outputs, and the in-tree runner supports the official scene-export route.

Method overview: Architecture Overview. VGGT- Ω appends camera and scene tokens (registers) to image tokens, and then alternates between global attention (or register attention) and frame attention layers. We replace the redundant.
Method overview: Architecture Overview. VGGT- Ω appends camera and scene tokens (registers) to image tokens, and then alternates between global attention (or register attention) and frame attention layers. We replace the redundant.

The recorded paper is arXiv 2605.15195.

Project page: vggt-omega.github.io. Code: facebookresearch/vggt-omega.

Build your run

Pipeline
VGGTOmegaPipelineworldfoundry.pipelines.vggt_omega.pipeline_vggt_omega
TaskInteractive Video interactive-video
Environmentworldfoundry-unified-cu128
DeviceCUDA 12.8
Runnerworldfoundry.pipeline
Loadingfrom_pretrained
Invocationnative_or_standard_invocation
worldfoundry-eval run \
  vggt-omega \
  --pipeline.task-profile interactive-video \
  --json

What it is for

vggt-omega is the recorded reconstruction line. Start here to emit geometry from the documented views.

It uses an official-architecture representation runner plus the existing WorldFoundry VGGTResult/save interface, emitting predictions.npz plus GLB scene artifacts. Weights stay under assets/checkpoints and are never committed into source files; the checkpoint license is CC-BY-NC-4.0. It is published by Meta AI.

Catalog tasks include 3d-reconstruction, gaussian-splatting, and novel-view-synthesis. Catalog aliases include vggt_omega and facebook/VGGT-Omega.

Not VGGT and VGGT-World.

Runtime contract

Optional input_path, interactions, camera_view, task_type, and image_resolution; artifacts scene_conf20.0_blackFalse_whiteFalse_camTrue_skyFalse_max200k.glb.

CLI task profiles: interactive-video. WorldFoundry binds VGGTOmegaPipeline.

Family status: Verified.

Use this entry for the verified official scene-export route with WorldFoundry representation artifacts. It is distinct from the vggt (base) and vggt-world entries — pick by route, not by name.

VGGT-Omega uses an official-architecture representation runner plus the existing WorldFoundry VGGTResult/save interface.

Running in WorldFoundry

Launch worldfoundry-eval run vggt-omega --pipeline.task-profile interactive-video to write geometry. The command builder above switches --pipeline.task-profile among recorded tasks and still emits worldfoundry-eval run.

It runs in the unified environment worldfoundry-unified-cu128 (Python 3.11, CUDA 12.8).

Recorded weight repositories include facebook/VGGT-Omega. Prepare the environment and check local assets with bash scripts/inference/prepare_model_infer.sh vggt-omega. Add --download to fetch missing public checkpoints.

Recorded licenses include cc-by-nc-4.0; status: Verified.