han-ol / InverseKinematicsSBI

Inverse Kinematics Benchmark for Simulation Based Inference

Inverse Kinematics Benchmark for Simulation-based Inference

Welcome to the repository of the inverse_kinematics_sbi package. TODO: package name

This is a lightweight library enabling you to flexibly specify a robot arm, compute the forward process, and solve the inverse kinematic problem.

Inverse kinematics was first proposed as a benchmark for simulation-based inference in the following paper:

Kruse, J., Ardizzone, L., Rother, C., & Köthe, U. (2021). Benchmarking Invertible Architectures on Inverse Problems (arXiv:2101.10763). arXiv. https://doi.org/10.48550/arXiv.2101.10763

(the code builds on https://github.com/vislearn/inn_toy_data/. TODO: not at the moment. mention related work appropriately)

Development

We manage dependencies in pyproject.toml and lock them in requirements.txt using pip-compile from the pip-tools suite. Automated formatting and checks are achieved using pre-commit.

Get started

1. Clone and enter the repository

git clone https://codeberg.org/han-ol/InverseKinematicsSBI.git && cd InverseKinematicsSBI

2. Create and activate an empty python environment, for example with conda

conda create -n ik-sbi python=3.10 && conda activate ik-sbi

3. Install dependencies with pip

pip install -r requirements.txt
pip install pre-commit  # TODO: as an optional dependency for development

4. Activate pre-commit hooks using

pre-commit install

TODO: Insert installation steps for pip-tools or optional dependency for development

Add a dependency

1. Add it to pyproject.toml
2. Run pip-compile
3. Install the updated requirements.txt using pip install -r requirements.txt