mmWave Base Stations with Cameras

Deep learning solutions are developed to tackle mmWave beam and link blockage predictions using camera feed. For more information, please refer to mmWave Base Stations with Cameras: Vision Aided Beam and Blockage Prediction.

Dependencies:

1. Python 3.7

2. Pytorch 1.3

3. NVIDIA GPU with a compatible CUDA toolkit (see NVIDIA website).

4. Processed ViWi dataset (see ViWi wesite)

Running the code:

The scripts available here are manily for training and testing a modified ResNet-18 model for mmWave beam prediction. With a little modification, they could also be used for the blockage prediction task. To train and test the model, you need to do the following:

1. Prepare two sets of data using the ViWi framework. One set is for training and the other is for testing (For more information on the data structure, see the next section).

2. Set the paths to the training and testing sets in the script "main.py" (i.e., modify train_dir and val_dir to point to you sets).

3. Set the path to where you want the trained network to be saved, by modifying net_name.

4. Run main.py

When the script finishes training and testing, it will save the accuracies in a result.mat file, and it will store the trained network.

Data Structure:

The script assumes a training and testing sets of data structured as a directory of subdirectories, as follows:

training_data
  |
  |- 1
  |- 2
  |- 3
  .
  .
  .
  |- x

where x = the size of the beam-forming codebook. The name of each sub-directory refers to the beam index in the codebook, and the contents of sub-directory "j", for example, are the images of those users served with the jth beam in the codebook.

Citation:

If you use this script or part of it, please cite the following:

@ARTICLE{2019arXiv191106255A,
       author = {{Alrabeiah}, Muhammad and {Hredzak}, Andrew and {Alkhateeb}, Ahmed},
        title = "{Millimeter Wave Base Stations with Cameras: Vision Aided Beam and Blockage Prediction}",
      journal = {arXiv e-prints},
     keywords = {Computer Science - Information Theory, Electrical Engineering and Systems Science - Signal Processing},
         year = "2019",
        month = "Nov",
          eid = {arXiv:1911.06255},
        pages = {arXiv:1911.06255},
archivePrefix = {arXiv},
       eprint = {1911.06255},
 primaryClass = {cs.IT},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2019arXiv191106255A},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

License:

This code package is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.