An Improved MangoNet Architecture Using Harris Hawks Optimization for Fruit Classification with Uncertainty Estimation

Abstract

Agriculture sector plays a crucial role in the development of the economic system for developing countries. Useful insights extracted from agricultural data would lead to a better economy in the upcoming future. It is hard to manually look for the ripened fruits for harvesting in a farmland of a big area. Automated image-based methods to identify the presence of fruits on a tree and determine the number of fruits present on the tree is a challenging problem. MangoNet architecture performs well in detecting the mangoes invariant. The proposed work improves the efficiency of the model by optimizing the dropout rate of the MangoNet architecture. Harris Hawks Swarm Optimization (HHO) algorithm is developed to tune the dropout rate of the model. The dropout rate is optimized using HHO in the architecture to avoid overfitting problem. The proposed method is compared with particle swarm optimization (PSO) and BAT algorithm (BA) which underestimate uncertainty. The simulation results exhibit improved accuracy over 80% when compared to other existing methods considered in this study. We show theoretically that this gives us the ability to capture uncertainty better than existing methods.