/IndusPose

Metal Parts’ Zero-Shot 6D Pose Estimation via Foundation Model and Template Update for Industrial Scenario

Primary LanguageJupyter Notebook

IndusPose

Metal Parts’ Zero-Shot 6D Pose Estimation via Foundation Model and Template Update for Industrial Scenario

Zero-shot Pose Estimation for Metal Parts

We provide the code for 6D pose estimation of novel metal parts using foundation model.

Given as inputs:

  • an RGB image,
  • the intrinsic parameters of the camera,
  • a mesh of the object,
  • a mask of that object in the image (bounding box maybe also ok)

Installation

It is east to install this code.

1. Clone the repository

The first step is to clone the repo and submodules:

git clone https://github.com/sunhan1997/IndusPose.git

2. Install the DepthAnything and GlueStick

pip install geomloss

3. Evaluation datasets:MP6D and ROBI

  • [MP6D]
  • [ROBI]
  • The scenes used in the paper for pose buffer and evaluation are shown in the DATASTE folder.

Run the demo

  • Render the template
python my_render.py
  • Estimate the metal parts' 6D pose
python run_pose.py

Results in the real wrold