Acies Vehicle Classifiers

Acoustic- and seismic-based vehicle classifiers.

If you use this repository in your research, please cite our accompanying paper to acknowledge the work. You can do so with the following BibTeX entry:

@inproceedings{li2024aciesos,
  title={Acies-OS: A Content-Centric Platform for Edge AI Twinning and Orchestration},
  author={Li, Jinyang and Chen, Yizhuo and Kimura, Tomoyoshi and et al.},
  booktitle={2024 33rd International Conference on Computer Communications and Networks (ICCCN)},
  pages={1--1},
  year={2024},
  organization={IEEE}
}

Setup

Python Environment Management

This project uses uv (or its predecessor rye) to manage the Python environment.

To check if rye is already installed, run:

which rye

• If the command prints a path, rye is available and you can skip this step.
• If not, we recommend installing uv for new setups. Follow uv’s official installation guide.

For backward compatibility, you may also install rye.

Clone and install dependencies

$ git clone [email protected]:acies-os/vehicle-classifier.git
$ cd vehicle-classifier
vehicle-classifier$ uv sync

# or, if using rye
vehicle-classifier$ rye sync

Install just

Install just use your package manager or pre-built binary.

Download Model Weights

Place the weight files under models/. You can either download them from the GitHub release or pull them directly with wget.

# in repo's root folder
vehicle-classifier$ cd models/

# you can also use curl, just replace `wget` with `curl -LO`
vehicle-classifier/models$ wget https://github.com/acies-os/vehicle-classifier/releases/download/weight-v1.0.0/gcq202410_mae.pt
vehicle-classifier/models$ wget https://github.com/acies-os/vehicle-classifier/releases/download/weight-v1.0.0/Parkland_TransformerV4_vehicle_classification_finetune_gcq202410_1.0_multiclasslatest.pt

Usage

There are 2 classifiers, to run them:

# VibroFM
just vfm

# FreqMAE
just mae

Documentation

The documentation is managed using Sphinx, which fetch docstring comments from code and compile them into html pages.

Sphinx sources live in the docs/source/ folder:

• conf.py — Sphinx configuration
• *.rst files — reStructuredText source documents that define docs content

To build Sphinx documentation, run this command:

just build-doc

Now the built documentation should live in docs/_build/ folder.

To view the documentation in browser, run this command:

just view-doc