The code implementation, and supplementary material, of the paper "Efficient Fingerprint Augmentation Evaluation on the Antwerp LoRaWAN Setting". Open access link to the paper: https://doi.org/10.5281/zenodo.13735399 If you use the material of this entry, please cite the current entry and the relevant paper as: Anagnostopoulos, G. (2024). Efficient Fingerprint Augmentation Evaluation on the Antwerp LoRaWAN Setting (Supplementary Material: Code). Zenodo. https://doi.org/10.5281/zenodo.13145395 Anagnostopoulos, G. (2024, September 9). Efficient Fingerprint Augmentation Evaluation on the Antwerp LoRaWAN Setting. International Conference on Indoor Positioning and Indoor Navigation (IPIN) 2024 (IPIN), Hong Kong. https://doi.org/10.5281/zenodo.13735399 ----------------------------------------------------------------------------------------------------------------------------------------------------------------- The following files are provided: Benchmarking_ProxyFAUG_LoRaWAN_Antwerp_v2_3.ipynb : The main Jupyter Notebook file of this work, including all the experiments of the paper. ProxyFAUG.py : This python script contains all methods implementing the ProxyFAUG augmentation scheme. (source: https://doi.org/10.5281/zenodo.4457353, check below for details) run_ProxyFAUG.py : This python script contains assisting methods that simplify the usage of ProxyFAUG in an experimental setting. More specifically, it provides functions that group actions, like creating new fingerprints using ProxyFAUG and then merging the new set with the original training sets, or grouping functions allowing to calculate and print performance statistics. haversine_script.py : Assisting script for geographic distance calculations, based on the Haversine project (https://pypi.org/project/haversine/). ----------------------------------------------------------------------------------------------------------------------------------------------------------------- The code implementation of the ProxyFAUG augmentation method is used as provided in the public record supplementing the original publication of the ProxyFAUG method. Therefore, the file ProxyFAUG.py, is taken from the original repository listed below, and the only change is the used dependencies that existed in imported libraries, have been removed. The sources are: Code: Anagnostopoulos, G., & Kalousis, A. (2021). ProxyFAUG: Proximity-based Fingerprint Augmentation (code). Zenodo, https://doi.org/10.5281/zenodo.4457353 Preprint: G. G. Anagnostopoulos and A. Kalousis, "ProxyFAUG: Proximity-based Fingerprint Augmentation," Arxiv, https://arxiv.org/abs/2102.02706 Paper: G. G. Anagnostopoulos and A. Kalousis, "ProxyFAUG: Proximity-based Fingerprint Augmentation," 2021 International Conference on Indoor Positioning and Indoor Navigation (IPIN), Lloret de Mar, Spain, 2021, pp. 1-7, doi: https://doi.org/10.1109/IPIN51156.2021.9662590 ----------------------------------------------------------------------------------------------------------------------------------------------------------------- The LoRaWAN dataset used in this work is a preprocessed version of the v1.3 of the following dataset: Data: Aernouts, M., Berkvens, R., Van Vlaenderen, K., & Weyn, M. (2019). Sigfox and LoRaWAN Datasets for Fingerprint Localization in Large Urban and Rural Areas (1.3) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.3904158 Data descriptor paper: Aernouts, M.; Berkvens, R.; Van Vlaenderen, K.; Weyn, M. Sigfox and LoRaWAN Datasets for Fingerprint Localization in Large Urban and Rural Areas. Data 2018, 3, 13. https://doi.org/10.3390/data3020013 In order to facilitate the reproducibility of our results as well as to enable consistent future comparisons, we reuse the same train/validation/test subsets, as they were split and publicized in our previous work: Data: Anagnostopoulos Grigorios, & Kalousis Alexandros. (2020). Analysing the data-driven approach of dynamically estimating positioning accuracy (data) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.4117818 Paper: G. G. Anagnostopoulos and A. Kalousis, "Analysing the Data-Driven Approach of Dynamically Estimating Positioning Accuracy," ICC 2021 - IEEE International Conference on Communications, Montreal, QC, Canada, 2021, pp. 1-7, doi: https://doi.org/10.1109/ICC42927.2021.9500369 Preprint: G. G. Anagnostopoulos and A. Kalousis, "Analysing the Data-Driven Approach of Dynamically Estimating Positioning Accuracy," Preprint, Arxiv, https://arxiv.org/abs/2011.10478 The same subsets introduced in https://doi.org/10.5281/zenodo.4117818 have been also used in another work: Code and Data: Anagnostopoulos Grigorios, & Kalousis Alexandros. (2022). Can I Trust This Location Estimate? Reproducibly Benchmarking the Methods of Dynamic Accuracy Estimation of Localization (code). Zenodo. https://doi.org/10.5281/zenodo.5589651 Paper: Anagnostopoulos, G.G.; Kalousis, A. Can I Trust This Location Estimate? Reproducibly Benchmarking the Methods of Dynamic Accuracy Estimation of Localization. Sensors 2022, 22, 1088. https://doi.org/10.3390/s22031088 In those works, apart from the train/validation/test subsets, another important operation took place, which is motivated and explained in the two above publications. We quote the text from https://doi.org/10.3390/s22031088 to justify the adjustment. "'Fingerprinting techniques are often compared to their counterpart, the ranging techniques such as multilateration, which require a minimum of three receiving gateways to produce a unique position estimate’ [12]. Even though satisfactory results can be obtained with fingerprinting methods when using messages with fewer than three receiving gateways, in our previous work [12], we reduced the dataset by only using the messages with at least three receiving gateways. A total of 75,054 messages with fewer than three receiving gateways were dropped, while 55,375 messages were retained to be used." We encourage future users of the LoRaWAN dataset v1.3, that want to work with messages having at least three receiving gateways, to reuse this train/validation/test subset split, to enhance the comparability of results with the works. The credit for the creation of these datasets goes to the creators of the primary dataset: Aernouts, M., Berkvens, R., Van Vlaenderen, K., & Weyn, M. In case you use the proposed subsets, please cite the Zenodo repository entry and the paper of the original creators, as well as the relevant Zenodo repository entry and the paper that introduced the subsets, as follows: Aernouts, M., Berkvens, R., Van Vlaenderen, K., & Weyn, M. (2019). Sigfox and LoRaWAN Datasets for Fingerprint Localization in Large Urban and Rural Areas (1.3) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.3904158 Aernouts, M.; Berkvens, R.; Van Vlaenderen, K.; Weyn, M. Sigfox and LoRaWAN Datasets for Fingerprint Localization in Large Urban and Rural Areas. Data 2018, 3, 13. https://doi.org/10.3390/data3020013 Anagnostopoulos Grigorios, & Kalousis Alexandros. (2020). Analysing the data-driven approach of dynamically estimating positioning accuracy (data) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.4117818 G. G. Anagnostopoulos and A. Kalousis, "Analysing the Data-Driven Approach of Dynamically Estimating Positioning Accuracy," ICC 2021 - IEEE International Conference on Communications, Montreal, QC, Canada, 2021, pp. 1-7, doi: https://doi.org/10.1109/ICC42927.2021.9500369