kngwyu / beach-cleaning-object-detection

Object detection model, and training data for use in a beach cleaning robot based in Okinawa, Japan

beach-cleaning-object-detection

Object detection model, and training data for use in a beach cleaning robot based in Okinawa, Japan

Install

Notebooks

In case you want to use conda virtual environment management, installation can be done as follows:

conda create --name beachbot
conda activate beachbot

pip install notebook *[or alternatively pip install jupyterlab]*
conda install -c anaconda ipykernel
python -m ipykernel install --user --name=beachbot

Alternatively, jupyter notebook can be installed via conda:

conda install -c conda-forge notebook
conda install -c conda-forge nb_conda_kernels

After this you should be able to start the notebooks like this:

cd [github repositor folder]
jupyter notebook

For kernel selection (in jupyter gui), you can select beachbot as python kernel. Dont forget to always do a conda activate beachbot before starting jupyter notebook in the git root directory, as special commands like % !pip install ... will be executed in the terminal environment!

Learned Models:

The learned models and model exports as ONNX, tensorflow & TensorRT are stored for now in Dropbox. To use them you have to:

1. Download this zip into the Models folder: https://www.dropbox.com/scl/fi/met65imjnito6x7wowe1g/beachbot_yolov5s_beach-cleaning-object-detection__v1i__yolov5pytorch.zip?rlkey=0vb4yt8ofppb08vj7imsp0vxc&dl=0
2. Extract the archive: unzip beachbot_yolov5s_beach-cleaning-object-detection__v1i__yolov5pytorch.zip

Datasets:

Download/export zip files of dataset beach-cleaning-object-detection.v1i.yolov5pytorch from Roboflow into Datasets folder