Scene Point Editing Toolkit (ScenePoint-ETK)

---

NOTE

This repository is under active update.

---

Scene Point Editing Toolkit (ScenePoint-ETK) is a python package to build various city scenes using Argoverse2 dataset

Paper

• arXiv link

Installation

# clone main repo
$ git clone https://github.com/ChadLin9596/ScenePoint-ETK

# create a python environment from (3.7 - 3.12) virtual environment
$ source <directory of virtual environment>/bin/activate
(your env)$ cd <this repository>
(your env)$ pip install .

# Note:
# it will automatically install py_utils from https://github.com/ChadLin9596/python_utils

# install `pptk` from https://github.com/ChadLin9596/pptk/releases
# for example (py3.9)
(your env)$ pip install https://github.com/ChadLin9596/pptk/releases/download/v0.1.1/pptk-0.1.1-cp39-none-manylinux_2_35_x86_64.whl

Example Usage

• unittest

  (your env)$ cd <this repository>
  (your env)$ python -m unittest

• tutorial jupyter-notebook under examples

Dataset Preparation

Please refer to Download SceneEdited Dataset page in the Documentation

Run ScenePoint-ETK/examples/01.setup_env_path.ipynb to assign their directories into this toolkit.

TODO

• complete tutorials at examples
  • basic usage
  • scalibitiy
  • trackability
  • portability
• website deployment
  • project page
  • documentation
• release v0 dataset
• more unittests

Citation

Please use the following citation for any work relies on the SceneEdited dataset or the ScenePoint-ETK toolkit.

@InProceedings{SceneEdited,
    author    = {Lin, Chun-Jung and Chin, Tat-Jun and Garg, Sourav and Dayoub, Feras},
    title     = {SceneEdited: A City-Scale Benchmark for 3D HD Map Updating via Image-Guided Change Detection},
    booktitle = {Proceedings of the Winter Conference on Applications of Computer Vision (WACV)},
    month     = {March},
    year      = {2026}
}