Classification and Grading of Okra-ladies finger using Deep Learning

Okra (Hibiscus esculentus) is a fruit/vegetable crop commonly known as lady's finger, Gumbo, Bhindi or gombo. The grading for Okra is performed to enable pricing equitableness. The different grades of the okra are based on the freshness, tenderness, color, shape, decay, scarred, bruised, cuts, insects, dirt, wormhole, and trim. In this paper, based on the length of the pod, four classifications performed are small, medium, large and extra-large. The machine vision technology is explored to grade the okra based on the length of the pod. Deep learning technology is emerging as the major approaches for signal and information processing with applications specific to image classification, speech recognition, and medical analysis. In grading the Okra's three deep learning models: AlexNet, GoogLeNet and ResNet50 are used. The dataset of ladies finger is collected, and the size of the dataset is 3200 of all the sizes. The accuracies obtained are 63.45% for AlexNet, 68.99% for GoogLeNet model and 99% for ResNet50 which is better than the others. The challenges are in the identification of the tenderness of the pods, wormholes, insects, and dirt.