Sani Depanshu ; Mahato Sandeep ; Saini Sourabh ; Agarwal Harsh Kumar ; Devshali Charu Chandra ; Anand Saket ; Arora Gaurav ; Jayaraman Thiagarajan

Abstract

The availability of well-curated datasets has driven the success of MachineLearning (ML) models. Despite greater access to earth observation data inagriculture, there is a scarcity of curated and labelled datasets, which limitsthe potential of its use in training ML models for remote sensing (RS) inagriculture. To this end, we introduce a first-of-its-kind dataset calledSICKLE, which constitutes a time-series of multi-resolution imagery from 3distinct satellites: Landsat-8, Sentinel-1 and Sentinel-2. Our datasetconstitutes multi-spectral, thermal and microwave sensors during January 2018 -March 2021 period. We construct each temporal sequence by considering thecropping practices followed by farmers primarily engaged in paddy cultivationin the Cauvery Delta region of Tamil Nadu, India; and annotate thecorresponding imagery with key cropping parameters at multiple resolutions(i.e. 3m, 10m and 30m). Our dataset comprises 2,370 season-wise samples from388 unique plots, having an average size of 0.38 acres, for classifying 21 croptypes across 4 districts in the Delta, which amounts to approximately 209,000satellite images. Out of the 2,370 samples, 351 paddy samples from 145 plotsare annotated with multiple crop parameters; such as the variety of paddy, itsgrowing season and productivity in terms of per-acre yields. Ours is also oneamong the first studies that consider the growing season activities pertinentto crop phenology (spans sowing, transplanting and harvesting dates) asparameters of interest. We benchmark SICKLE on three tasks: crop type, cropphenology (sowing, transplanting, harvesting), and yield prediction