fachu000 / deep-autoencoders-cartography

This repository contains the code reproducing the results in the paper "Deep Completion Autoencoders for Radio Map Estimation"

Introduction

The Python code in python folder implements the simulations and plots the figures described in the paper "Deep Completion Autoencoders for Radio Map Estimation" by Yves Teganya and Daniel Romero.

Required packages

Use the package manager pip to install the following packages:

tensorflow
scipy
cvxpy
cvxopt
matplotlib
pandas
joblib
sklearn
opencv-python

Guidelines

Add the simulation environment for Python with the following git command:

git submodule add https://github.com/fachu000/GSim-Python.git ./gsim

The first time one wants to run a simulation after downloading the required Python packages and the simulation environment, one enters the folder python and creates the folder output with subfolder autoencoder_experiments. Inside the latter, create two subfolders savedWeights and savedResults. After that, one will be able to execute any simulation in the aforementioned paper by running run_experiment.py file that is located in the folder python/ followed by the experiment number as will be explained with an example later.

The experiments reproducing different figures in the paper are organized in methods located in the file Experiments/autoencoder_experiments.py. The comments before each method indicate which figure(s) on the paper it generates.

For experiments that use the Wireless Insite software, please unzip the dataset files from remcom_maps.zip in the Generators folder to form the remcom_maps folder.

One is now all set. For example, to run experiment 1003, one types run_experiment.py 1003. To just display the results of the last execution of experiment 1003 (stored in output/autoencoder_experiments), one types run_experiment.py -p 1003. Note that this way of only plotting executed experiments applies for displaying 1D curves, it does not work for 2D images. The simulation results for experiments that produce 2D images (possibly together with 1D curves) are saved in output/autoencoder_experiments/savedResults.

For any questions related to the code or difficulties to run it, please send me an email at [e-mail] or [e-mail]

Citation

If our code is helpful in your resarch or work, please cite our paper.

@article{teganya2020deepcompletion,
  title={Deep Completion Autoencoders for Radio Map Estimation},
  author={Teganya, Yves and Romero, Daniel},
  journal={arXiv preprint arXiv:2005.05964},
  year={2020}
}