AUTHOR=Sharma Arun , Satish Deepshikha , Sharma Sushmita , Gupta Dinesh TITLE=iRSVPred: A Web Server for Artificial Intelligence Based Prediction of Major Basmati Paddy Seed Varieties JOURNAL=Frontiers in Plant Science VOLUME=10 YEAR=2020 URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2019.01791 DOI=10.3389/fpls.2019.01791 ISSN=1664-462X ABSTRACT=

The purity of seeds is the most important factor in agriculture that determines crop yield, price, and quality. Rice is a major staple food consumed in different forms globally. The identification of high yielding and good quality paddy seeds is a challenging job and mainly dependent on expensive molecular techniques. The practical and day-to-day usage of the molecular-laboratory based techniques are very costly and time-consuming, and involves several logistical issues too. Moreover, such techniques are not easily accessible to paddy farmers. Thus, there is an unmet need to develop alternative, easily accessible and rapid methods for correct identification of paddy seed varieties, especially of commercial importance. We have developed iRSVPred, deep learning based on seed images, for the identification and differentiation of ten major varieties of basmati rice namely, Pusa basmati 1121 (1121), Pusa basmati 1509 (1509), Pusa basmati 1637 (1637), salt-tolerant basmati rice variety CSR 30 (CSR-30), Dehradoon basmati Type-3 (DHBT-3), Pusa Basmati-1 (PB-1), Pusa Basmati-6 (PB-6), Basmati -370 (BAS-370), Pusa Basmati 1718 (1718) and Pusa Basmati 1728 (1728). The method has an overall accuracy of 100% and 97% on the training set (total 61,632 images) and internal validation set (total 15,408 images), respectively. Furthermore, accuracies of greater than or equal to 80% have been achieved for all the ten varieties on the external validation dataset (642 images) used in the study. The iRSVPred web-server is freely available at http://14.139.62.220/rice/.