Depth Anything V1

Depth Anything V1 from TikTok / The University of Hong Kong: depth and geometry; Verified.
Depth Anything V1 is the foundation model for relative monocular depth that popularized large-scale unlabeled-data depth training. In this catalog both integration and runner status are verified: the in-tree pipeline exports relative depth and geometry artifacts, and the Studio route produced a visually validated nonblank depth map on 2026-07-10.

The recorded paper is arXiv 2401.10891.
Project page: depth-anything.github.io. Code: LiheYoung/Depth-Anything.
The official paper is on arXiv 2401.10891. Project: depth-anything.github.io. Code: LiheYoung/Depth-Anything. Weights: LiheYoung/depth-anything-small-hf, base-hf, large-hf. Publisher: TikTok / HKU. Catalog license is unset.
This card is not Depth Anything V2 prior (metric) and not DAP (360° panoramas).
Build your run
DepthAnything1Pipelineworldfoundry.pipelines.depth_anything.pipeline_depth_anything_v1defaultworldfoundry.pipelineworldfoundry-eval run \
depth-anything-v1 \
--pipeline.task-profile default \
--jsonWhat it is for
Pick depth-anything-v1 for feed-forward geometry: views in, reconstruction / trajectory / point cloud out. Use it when you need the catalog 3D route rather than a depth-only prior.
It is published by TikTok / The University of Hong Kong.
Catalog tasks include depth, geometry, and monocular-depth-estimation. Documented modalities: image in, depth out.
Runtime contract
Optional input_path, grayscale, and operator_kwargs; artifacts depth-anything-v1.png.
CLI task profiles: default. WorldFoundry binds DepthAnything1Pipeline.
Family status: Verified.
A verified, dependable prior — good default for relative depth. For metric depth choose depth-anything-v2-prior or depth-anything-v3-prior instead.
In-tree Depth Anything V1 pipeline is registered for relative depth and geometry artifact export.
Running in WorldFoundry
Launch worldfoundry-eval run depth-anything-v1 --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 LiheYoung/depth-anything-large-hf. Prepare the environment and check local assets with bash scripts/inference/prepare_model_infer.sh depth-anything-v1. Add --download to fetch missing public checkpoints.
Status: Verified.
