Bachelor Thesis: Convolutional Neural Networks with Local Context Masks

This work proposes, implements and evaluates modified convolution operation within CNNs which explicitly considers additional context when processing data. This is done through a mechanism where local context corresponds to a local real-valued convolution mask which assigns higher weights to contextually close data and lower weights to contextually distant data.

Download paper here

BibTeX:

@inproceedings{Paplhm2020ConvolutionalNN,
  title={Convolutional Neural Networks with Local Context Masks},
  author={Jakub Paplh{\'a}m},
  year={2020},
  url={https://api.semanticscholar.org/CorpusID:221709465}
}