Second Master's Research Project for Leiden University, focused on attempting to use machine learning to identify radio sources and optical counterparts in LOFAR data
The easiest way to install this package is with pip with pip install lofarnn.
Otherwise, the lastest code can be built with pip install git+https://github.com//jacobbieker/lofarnn.git
The different PyTorch models and datasets can be easily imported from the lofarnn package. To preprocess LOFAR data into the correct format for either CNN or Detectron2 models, example code can be found under examples folder.
PyTorch models used in the thesis are available here: https://drive.google.com/drive/folders/1lCFcQT7WRTiMxfd8jL2ReCoJrNAhj4BW?usp=sharing. The best performing model is the multi_cnn.pth model.