Traffic Control Gesture Recognition for Autonomous Vehicles

This is the official project page including paper link, code, baseline models and dataset.

Traffic Control Gesture Recognition for Autonomous Vehicles
Julian Wiederer*, Arij Bouazizi*, Ulrich Kressel, Vasileios Belagiannis
Accepted for oral presentation at IROS 2020 oral
*denotes equal contribution

Requirements

List of packages used to train and test the models (see also requirements.txt).

h5py==2.10.0
tensorflow==2.0.0
Keras==2.3.1
matplotlib==3.3.1
numpy==1.17.2
tqdm==4.48.2
keras_self_attention==0.42.0
scikit_learn==0.23.2

Dataset

Requesting Access

To optain the TCG dataset (around 650 MB), please send an email to tcg.recognition@gmail.com with subject matter '[IROS2020] TCG Dataset - Access Request' stating the following:

       Your name, title and affiliation
       Your intended use of the data
       The following license statement:
       With this e-mail I agree that I will use the TCG dataset following the Creative Common Attribution-NonCommercial 4.0 International License.

We will promptly reply with a download link. Download the .zip file and copy it into dataset. Unzip using

cd dataset
unzip tcg_dataset.zip

License

TCG dataset is licensed under Creative Commons Attribution-NonCommercial 4.0 International License.

Content

The dataset includes two files, the description file with all required meta information and annotations tcg.json and the data file tcg_data.npy with the 3D pose sequences.

Baseline Models

Download the baseline models from google drive and unzip the .zip file in the baselines directory:

https://drive.google.com/file/d/1Kz-7PpSkucCSBWcrL9K3fTotxhNgWAaa/view?usp=sharing

Training

Make sure the requirements are satisfied in your environment, and the dataset is downloaded. Change configuration and hyperparameters in /config/config.py as you like and run

python config.py

to initialize config.ini. Then run

python main.py 

You can also play with the hyperparameters using the /config/config.py.

Evaluation

To reproduce the quantitative results in our paper, run the provided script:

python evaluate.py

Citation

We believe in open research and we will be happy if you find this dataset usefull for your work. If you use it, please consider citing our work.

@InProceedings{wiederer2020traffic,
               title={Traffic Control Gesture Recognition for Autonomous Vehicles}, 
               author={Julian Wiederer and Arij Bouazizi and Ulrich Kressel and Vasileios Belagiannis},
               year={2020},
               booktitle={ArXiv preprint}
}

Contact

If you have any questions feel free to contact julian.wiederer@uni-ulm.de.