Deep Neural Network Based Yield Phenotyping Framework: Fruit Grading and Classification

This research paper instigates an Artificial Intelligence-based Yield Phenotyping Framework tuned for fruit classification and grading. Fruit species exhibit homogeneity in color, shape, and size qualities, resulting in fruit grading and segmentation tasks on such datasets being counterproductive and vulnerable. The primary objective of this research paper is to establish a real-time, non-destructive system that automates fruit segmentation, grading and classification. Here, we use two different datasets and conduct a detailed analysis of the architectures of two deep learning models: AlexNet and ResNet. Different kinds of fruits are added to the proposed prototype. Videos of each fruit is recorded and sent to a local server. Deep learning models predict the fruit grade, and the system's conveyor belt sort the fruits into baskets based on their grade. By using preprocessing techniques, we can precisely isolate the fruit in each image. This study considers evaluation metrics such as accuracy, precision, recall and f1-score to evaluate the deep learning algorithm models. The ResNet architecture enables research to achieve a peak accuracy, precision and recall of 99%. This paper presents research surrounding the proposed prototype's architecture, methodologies, and experimental results and therefore provides an automated solution to an existing challenge in the Agri-field management.