csv and pytorch tensors

Data associated with the article: "PM2.5 Retrieval with Sentinel-5P Data over Europe Exploiting Deep Learning"

This dataset provides pre-processed Sentinel-5P imagery reprojected onto the CAMS grid for PM2.5 estimation. Each sample contains the first 60 principal components extracted from the Sentinel-5P spectral bands, excluding the UV range, after applying Principal Component Analysis (PCA). The final band in each sample represents the PM2.5 concentration values obtained from the CAMS dataset.

Multi-label event classification label of each sample-document is done with nine bits. The first bit signifies whether an event is present or absent with 1 or 0 respectively. The remaining eight bits signifies presence or absence of (i) covid, (ii) flood, (iii) storm, (iv) heavy rain, (v) cloudburst, (vi) landslide, (vii) earthquake, (viii) Tsunami with 1 or 0. The location and the impact sentence classification labeling are similar.