A python toolbox based on PyTorch which utilized neural network for rain estimation and classification from commercial microwave link (CMLs) data. This toolbox provides an implementation of algorithms for extracting rain-rate using neural networks and CMLs. Addinaly this project provides an example dataset with data from two CMLs and implementation of performance and robustness metrics.
This project is part of the M.Sc. thesis of Hai Victor Habi, under the supervision of Prof. Hagit Messer at Tel Aviv University.
PyNNcml is available on PyPI. To install, run the following command:
pip install pynncml
| Python Version | Status |
|---|---|
| Python 3.8 |  |
| Python 3.9 |  |
| Python 3.10 |  |
| Python 3.11 |  |
- Wet Dry Classification (RNN[1,2,3] and STD Window[6])
- Rain Estimation (Constant Baseline [6], Dynamic Baseline [5], Direct RNN Estimation [4,3])
- Rain Field Generation (RainGAN )
- Rain Field Interpolation (IDW, GMZ [9])
This repository includes an example of a dataset with a reference rain gauge. In addition, this repository provide PyTorch version of the OpenMRG dataset [10].
This repository includes examples of how to use the PyNNcml library to obtain various whether monotoning information using CMLs data. Please see the examples directory for more information.
In this project we supply a set of trained networks in our Model Zoo, this networks are trained on our own dataset which is not publicly available. The model contains three types of networks: Wet-dry classification network, one-step network (rain estimation only) and two-step network (rain estimation and wet-dry classification). Moreover, we have provided all of these networks with a various number of RNN cells (1, 2, 3). From more details about network structure and results see the publication list.
PyNNCMLs aims at keeping a more up-to-date fork and welcomes contributions from anyone.
If you would like to contribute, please follow the contribution guidelines.
Please cite one of following paper if you found our neural network model useful. Thanks!
[1] Habi, Hai Victor and Messer, Hagit. "Wet-Dry Classification Using LSTM and Commercial Microwave Links"
[2] Habi, Hai Victor and Messer, Hagit. "RNN MODELS FOR RAIN DETECTION"
[3] Habi, Hai Victor. "Rain Detection and Estimation Using Recurrent Neural Network and Commercial Microwave Links"
[4] Habi, Hai Victor, and Hagit Messer. "Recurrent neural network for rain estimation using commercial microwave links." IEEE Transactions on Geoscience and Remote Sensing 59.5 (2020): 3672-3681.
Also, this package contains the implementations of the following papers:
[5] J. Ostrometzky and H. Messer, “Dynamic determination of the baselinelevel in microwave links for rain monitoring from minimum attenuationvalues,”IEEE Journal of Selected Topics in Applied Earth Observationsand Remote Sensing, vol. 11, no. 1, pp. 24–33, Jan 2018.
[6] M. Schleiss and A. Berne, “Identification of dry and rainy periods using telecommunication microwave links,”IEEE Geoscience and RemoteSensing Letters, vol. 7, no. 3, pp. 611–615, 2010
[7] Jonatan Ostrometzky, Adam Eshel, Pinhas Alpert, and Hagit Messer. Induced bias in attenuation measurements taken from commercial microwave links. In 2017 IEEE International
Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 3744–3748. IEEE,2017.
[8] Jonatan Ostrometzky, Roi Raich, Adam Eshel, and Hagit Messer. Calibration of the attenuation-rain rate power-law parameters using measurements from commercial microwave networks. In 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 3736–3740. IEEE, 2016.
[9] Goldshtein, Oren, Hagit Messer, and Artem Zinevich. "Rain rate estimation using measurements from commercial telecommunications links." IEEE Transactions on signal processing 57.4 (2009): 1616-1625.
And include PyTorch implementation of the OpenMRG dataset:
[10] van de Beek, Remco CZ, et al. OpenMRG: Open data from Microwave links, Radar, and Gauges for rainfall quantification in Gothenburg, Sweden. No. EGU23-14295. Copernicus Meetings, 2023.
If you found one of those methods usefully please cite.