This repository is the official implementation of [Convolutional Neural Networks with Local Context Masks].
PyTorch 1.5.0
Python 3.7.2
CUDA 10.0.130
For comparison with Pixel-Adaptive Convolution, pacnet must be cloned inside the repository.
The training framework is implemented for use within RCI cluster data storage. User must first edit data_loader/data_loaders.py and configuration files to begin training elsewhere.
To train the models using method described in the thesis, run this command:
python train.py -c <path_to_config> -r <path_to_model_checkpoint>