Dataset for GLOBECOM paper "Spectrum Monitoring for Radar Bands using Deep Convolutional Neural Networks"

This dataset refers to all the images+labels used for training a Convolutional Neural Network (CNN) for RF signal classification/monitoring. All samples in this dataset were collected over-the-air using a USRP N210 RF front-end. The raw IQ samples were converted into two different representations: (i) spectrograms, and (ii) amplitude+phase shift matrices. Both representations were then stored in JPEG format. The label files are present in txt format, with each line defining the image path and its class (0 or 1). Class 1 means that radar emissions (the incumbent) are present in the image, while class 0 means that they are not. Several types of radar signal waveforms were stored (e.g. LFM, barker, and, pulsed carrier). Class 0 means that no radar emissions are present. In both class 0 and class 1 images, WiFi and LTE signals may be present. For more information regarding the context/scenario where the CNN was used, we encourage reading the paper https://arxiv.org/pdf/1705.00462.pdf