T4 devkit

A toolkit to load and operate T4 dataset.

Getting started

📘 Documentation | ⚙️ Tutorial | 🧰 API Reference

Installation

Install via GitHub

Note that the following command installs the latest main branch:

# e.g) with pip
pip install git+https://github.com/tier4/t4-devkit.git

By specifying @<TAG_OR_BRANCH>, you can install the particular version of t4-devkit:

# e.g) with pip
pip install git+https://github.com/tier4/t4-devkit.git@main

Install from source

You need to install uv. For details, please refer to OFFICIAL DOCUMENT.

git clone git@github.com:tier4/t4-devkit.git
cd t4-devkit
uv sync --python 3.10

The virtual environment can be activated with:

source .venv/bin/activate

Dataset Format

For details of T4 dataset format, please refer to tier4_perception_dataset/t4_format_3d_detailed.md.

Feature supports

Visualization

Feature	Task	Support
3D	3D Boxes	✅
	PointCloud Segmentation
	Raw PointCloud	✅
	3D Trajectories	✅
	TF Links	✅
2D	2D Boxes	✅
	Image Segmentation	✅
	Raw Image	✅
	Raw PointCloud on Image	✅
Map	Vector Map
	Ego Position on Street View	✅

Evaluation

Feature	Task	Support
3D	3D Detection
	3D Tracking
	3D Motion Prediction
	PointCloud Segmentation
2D	2D Detection
	2D Tracking
	Image Segmentation
	Classification