%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Dataset for the CTW Competition 2019: By Maximilian Arnold, Marc Gauger, and Stephan ten Brink. Contact: arnold / gauger / tenbrink@inue.uni-stuttgart.de This code and dataset is provided for the CTW Competition: Indoor user localisation. It is licensed under the Creative Commons Attribution 4.0 license. If you in any way use this code for research that results in publications, please cite it appropriately. Paper: [1] Novel Massive MIMO Channel Sounding Data Applied to Deep Learning-based Indoor Positioning, SCC 2019 (available on arxiv) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Required software environment to run example notebook: - Python, Keras, Numpy, Jupiter (Best to start at https://colab.research.google.com/) - Please note that we cannot offer a comprehensive tech support, but google and tensorflow should be fine to get you going %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% How to get started: - The description of the Competition and Dataset is in the Folder "0_CompetitionDescription"; Please read carefully. If you are interested in how to measurement was done see [1]. - Some measurement sanity checks and first predictions from the INUE are shown in 2_SanityChecks"; Take a look! - There is a Notebook in the main directory called "CTW2019_StartingNotebook.ipynb" (a simple dense net is provided to get you started; you may change anything you like, e.g., data format for feeding the network, net structure, etc...) - The datasets are available in different formats (The jupyter takes Matload from Hdf5.loadmat function): MAT: https://drive.google.com/open?id=1zMHtrTjr9Uf08Jh2yo24ufsyi-9iQfFD NP: https://drive.google.com/open?id=1Uv4Ys-jJCtsiBgl5w7YYASyAmh8_lT7d Tere is a position and channel mat file + SNR file. Be careful with dimensions, Antennas there are 16, subcarriers 924, train is (17486 long). 2000 for test. Use print shape to be sure. the first index shows the order for each channel to each position. 16 antennas were measured and 924 subcarriers used; COMPLEX Coefficients please be careful! %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% How to compete: In the morning of May 26.5.2019, at IEEE CTW, a set of test data will be distributed. Participating teams may run their algorithms on this test data set, and submit their estimated position coordinates to data-competition@inue.uni-stuttgart.de no later than 20:00. Winners will be determined by the organizers by evaluating and comparing the root-mean-square position error. (Ground truth will not be available to the participants, and the test dataset will not be available before the conference. It will be averaged over all test positions) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Now get started and we wish you the best luck and enough coffee!