Deep Learning Models for Classification of Okra Fruit Diseases
Harshit Dheeraj, Akshatha Prabhu, N. Shobha Rani, B Jeevan
Abstract
Agriculture is a major industry in India and must continually improve if it is to keep its competitive advantage in the world market. Major of the crops produced are susceptible to loss in production due to the failure in detection and diagnose of diseases, lack of expertise, inadequate control measures. In this paper, different defects on Okra are identified – blackspots, fungus and rotten. The healthy images are also considered along with the defected samples. Convolution neural networks – VGG 16, Resnet 50 and Resnet 101 are used. Total of 1,913 images of Okra are collected, with a train-test ratio of 80:20. The accuracies of VGG -16, Resnet 50 and Resnet101 achieved are 98.63%, 92.47 % and 95.89 % respectively. The models have been trained for various values of epochs. The Precision, Recall and F1- score have also been measured across all the models.