KishanRakshak : A Transfer Learning Approach to Classification and Prediction of Rice Crop Damage Estimation in India for Effective Insurance Claims

Abstract

For countries where primary sector of economy is agriculture, the claim for insurance based on crop damage is a common phenomenon. To make the related processes like claim assessment, faster disbursement etc. more effective and faster, it is essential to have a proper damage assessment of the crop fields. KishanRakshak is a transfer learning approach based Convolutional Neural Network(CNN) model which when applied and fine-tuned on a custom made dataset classified the percentage of damage that has occurred in the field. These classifications are adhering to the government rules. Instead of making use of drones to capture the images of damaged crops which is rather a costly process, images are obtained through smartphones' cameras at certain angles making it much cost effective. On experimentation conducted over available as well as custom made datasets the proposed model has achieved a classification accuracy of 94.67 %. KishaRakshak, is a novel and productive approach to facilitate farmers in India with easier insurance claim assessment as well as disbursement.