WorldFoundry
English中文

Depth Anything V2

Depth Anything V2

Depth Anything V2 from TikTok / The University of Hong Kong: monocular-depth-estimation and metric-depth-estimation; Verified.

Edit this pageGitHubMarkdown

Depth Anything V2 is the sharper, synthetic-data-trained successor to V1, integrated here as a metric/monocular depth prior with verified runner status. The recorded checkpoint is depth-anything/Depth-Anything-V2-Large; the shared priors factory and the video-depth prior reuse this implementation.

Official paper teaser: Depth Anything V2 recovers finer structure than V1 and SD-based depth models.
Official paper teaser: Depth Anything V2 recovers finer structure than V1 and SD-based depth models.

The recorded paper is arXiv 2406.09414.

Project page: depth-anything-v2.github.io. Code: DepthAnything/Depth-Anything-V2.

The official paper is on arXiv 2406.09414. Project: depth-anything-v2.github.io. Recorded checkpoint: depth-anything/Depth-Anything-V2-Large (depth_anything_v2_vitl.pth). Runner is verified through the shared prior contract. Per-run evidence stays outside the catalog.

Build your run

Pipeline
DepthAnythingV2PriorPipelineworldfoundry.pipelines.geometry_priors.pipeline_geometry_priors
TaskDefault default
Environmentworldfoundry-unified-cu128
DeviceCUDA 12.8
Runnerworldfoundry.pipeline
Loadingfrom_pretrained
Invocationnative_or_standard_invocation
worldfoundry-eval run \
  depth-anything-v2-prior \
  --pipeline.task-profile default \
  --json

What it is for

Use depth-anything-v2-prior for catalog geometry inference. Choose it when the artifact is a reconstruction, not a generated video.

It is published by TikTok / The University of Hong Kong.

Catalog tasks include monocular-depth-estimation, metric-depth-estimation, and geometry-prior. Catalog aliases include dav2-prior and depth-anything-v2. Documented modalities: image in, depth out.

Runtime contract

Optional input_path, model, execute, max_points, and focal_length; artifacts depth-anything-v2-prior.npz.

CLI task profiles: default. WorldFoundry binds DepthAnythingV2PriorPipeline.

Family status: Verified.

It is integrated as an independent geometry prior under worldfoundry/base_models and exposed through the shared geometry-prior pipeline/operator contract. Higher-level pipelines (e.g. ViPE) import it from this canonical owner instead of keeping private copies. Runner parity is verified through the shared contract.

Integrated as an independent depth prior at worldfoundry/base_models/three_dimensions/depth/depth_anything/depth_anything_v2.

Running in WorldFoundry

Launch worldfoundry-eval run depth-anything-v2-prior --pipeline.task-profile default 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 depth-anything/Depth-Anything-V2-Large. Prepare the environment and check local assets with bash scripts/inference/prepare_model_infer.sh depth-anything-v2-prior. Add --download to fetch missing public checkpoints.

Status: Verified.