❄️ SnowScooterDetector - THIS REPOSITORY IS DEPRECATED. IF YOU WANT TO RUN THE SNOWMOBILE MODEL ON YOUR DATA PLEASE REFER TO THE FOLLOWING REPO: https://github.com/NINAnor/snowmobile_analyzer
Please make sure that the correct parameters are set on the config_training.yaml if you are going to train a model or config_inference if you are planning to use the model to detect snowscooters.
This code has been tested using Ubuntu 18.04 LTS and Windows 10 but should work with other distributions as well. Only Python 3.8 is supported though the code should work with other distributions as well.
- Clone the repository:
git clone https://github.com/NINAnor/snowscooter_detector
- Install requirements:
We use poetry as a package manager which can be installed with the instructions below:
cd snowscooter_detector
pip install poetry
poetry install --no-root
- Pydub and Librosa require audio backend (FFMPEG)
sudo apt-get install ffmpeg
First you need to have docker installed on your machine. Please follow the guidelines from the official documentation.
- Clone the repository:
git clone https://github.com/NINAnor/snowscooter_detector
- Build the image
cd snowscooter_detector
docker build -t SnowscooterDet -f Dockerfile .
After updating the parameters in config_training.yaml run the training script:
poetry run python training/lightning_trainer/train_pipeline.py
Or alternatively, if you have docker installed and the docker image built:
docker run --rm -v $PWD/:/app -v /PATH/TO/YOUR/DATA/:/Data SnowscooterDet python training/lightning_trainer/train_pipeline.py
💡 If you are using Docker, don't forget the flag --nv to expose your GPU 💡
After updating the parameters in config_inference.yaml run the prediction script:
poetry run python inference/predict.py
Or alternatively, if you have docker installed and the docker image built:
docker run --rm -v $PWD/:/app -v /PATH/TO/YOUR/DATA/:/Data SnowscooterDet python inference/predict.py
💡 If you are using Docker, don't forget the flag --nv to expose your GPU 💡
⭐ Please note our scripts support PyfileSystem. This means that it is possible to add a CONNECTION_STRING argument in the form user:password@host in config_inference or config_training for the scripts to fetch the data on your remote server.
⭐ If your data is stored locally simply set CONNECTION_STRING to False.
If you come across any issues with any of our scripts, please open an issue on the project's GitHub.
For other inquiry you can contact me at [email protected].
This project is the result of a collaboration between NINA and SINTEF and has been funded by the Norwegian Research Council.