depthanything3-local

Run Depth Anything 3 inference locally for depth and camera pose estimation.

Updated May 7, 2026
One-click install
npx skills add https://github.com/EurecaMoment/BenchClaw --skill depthanything3-local
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: depthanything3-local
Source: https://github.com/EurecaMoment/BenchClaw/tree/main/BenchClaw/annotation-tool/depthanything3
Command: npx skills add https://github.com/EurecaMoment/BenchClaw --skill depthanything3-local

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires Depth-Anything-3, and includes scripts (resource) and assets (resource) components.

What problem does it solve?

This Skill provides a local service for Depth Anything 3 (DA3) inference on various image and video formats, enabling tasks like depth map creation, camera pose estimation, and scene export.

Core Features & Use Cases

  • Depth Estimation: Perform depth estimation from images, folders, videos, or COLMAP-style inputs.
  • Camera Pose Estimation: Estimate camera poses for video frame processing.
  • Scene Export: Export 3D scene models with GLB format and depth visualizations.
  • Use Case: For a researcher analyzing 3D scenes from videos, use this Skill to perform depth estimation and pose estimation, followed by exporting the scene model for further analysis.

Quick Start

Execute depth inference on the provided image 'scene.png' and export results to 'output':

depthanything3-client.py auto --input-path scene.png --export-dir output

Frequently Asked Questions about depthanything3-local

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I perform local depth estimation on video frames and export a 3D scene?

To perform local depth estimation on video frames and export a 3D scene, use a local Depth Anything 3 inference service to process the video, estimate camera poses, and export the resulting 3D scene model in GLB format.

What is COLMAP input processing in the context of 3D depth estimation?

COLMAP input processing for 3D depth estimation involves feeding COLMAP-style image datasets into Depth Anything 3 to generate detailed depth maps and camera pose estimations for complex scene reconstruction.

Do I need a Python environment and FastAPI backend installed to use Depth Anything 3 locally?

Yes, running Depth Anything 3 locally requires a locally installed DA3 backend and a Python environment to support the FastAPI backend for handling image, video, and COLMAP processing requests.

Can I estimate camera poses from a folder of images for 3D modeling?

Yes, you can estimate camera poses from a folder of images for 3D modeling by processing the directory through Depth Anything 3, which supports depth estimation and pose tracking for various input formats.

What is the best way to run depth inference on an image and save the depth visualization?

The best way to run depth inference on an image and save the depth visualization is to execute the Depth Anything 3 client script with your input path and an export directory to output the 3D scene model and visualizations.

Why does local Depth Anything 3 inference require a FastAPI backend?

Local Depth Anything 3 inference requires a FastAPI backend to provide a local service interface that efficiently handles and routes inference requests for depth map creation and scene export across various media formats.