This software produces a graph representation derived from point cloud data, which is then used as input for a Graph Neural Network (GNN).This allows to increase the amount of data by the factor of 1000.
See some examples of usage in bird-cloud-gnn-experiments repo.
In scenarios where labeled data is limited, there's a pressing need to expand the dataset effectively. One effective strategy involves altering the data's representation. In this context, we adopted such an approach by acquiring a graph representation from point cloud data. Depending on the chosen parameters, this transformation can augment the dataset by a factor of up to 1000. Subsequently, this graph representation is harnessed as input for Graph Neural Networks (GNNs). GNNs are highly sought after due to their innate ability to adeptly capture and leverage the inherent properties of graph-structured data. They excel in modeling intricate network relationships, autonomously acquiring informative features, and facilitating effective knowledge transfer.
To install bird_cloud_gnn from GitHub repository, do:
git clone https://github.com/point-cloud-radar/bird-cloud-gnn.git
cd bird-cloud-gnn
python3 -m pip install .The documentation can be found on Read the Docs.
If you want to contribute to the development of bird_cloud_gnn, have a look at the contribution guidelines.
This package was created with Cookiecutter and the NLeSC/python-template.