Software Setup

This page describes how to install the TRumi dataset generation pipeline on a Linux PC.

Prerequisites

• Python 3.12

• uv package manager

• Docker (required for ORB-SLAM3)

• ExifTool

On Ubuntu, install ExifTool with:

sudo apt install libimage-exiftool-perl

Install the Package

Clone the repository and install the package:

cd ~
git clone https://github.com/TrossenRobotics/trumi.git
cd trumi
uv sync

This installs all dependencies declared in pyproject.toml into a local .venv.

To activate the environment manually:

source .venv/bin/activate

Or prefix any command with uv run to automatically use the environment without activating it.

Build the ORB-SLAM3 Docker Image

The SLAM pipeline runs ORB-SLAM3 inside a Docker container. Build the image once before running the pipeline.

Follow the setup instructions in TrossenRobotics/ORB_SLAM3 — DOCKER.md.

Verify Docker is running before you use the pipeline:

docker info

Developer Setup (Optional)

If you plan to contribute, install the pre-commit hooks:

uv run pre-commit install

Try It Without Recording (Optional)

An example dataset is available on Hugging Face so you can try the pipeline before recording your own data.

Download it with the Hugging Face CLI (downloads into the current directory, preserving the example_trumi_dataset/ folder structure):

cd ~/trumi
hf download TrossenRoboticsCommunity/example_trumi_dataset \
    --repo-type dataset \
    --local-dir .

With the software installed, continue to Data Collection to record a session, or jump straight to Dataset Generation Pipeline if you are using the example dataset.