Convolutional Neural Network based Rotten Fruit Detection using ResNet50

Food contains essential nutrients for human beings to grow and develop. Out of so many type of food, vegetables and fruits are important for humans’ daily healthy diet as they provide all the nutrients that helps human to prevent diseases. However, fruit will get rotten easily if not store properly due to the spread of bacteria. Therefore, it is important for food industry to perform inspection on fruits before selling to the consumers. The problem encountered in the human inspection is lower in consistency and accuracy as the manual inspection by humans’ eye will consume time and energy. To solve this problem, the proposed method is to apply the deep learning technique which is Convolutional Neural Networks (CNNs) for feature extraction and classification of rotten fruits. The types of fruits that will be detected and classified in this paper are banana, apple and orange. The validation accuracy obtained in this paper is 98.89%. The total duration of training stage is 212.13 minutes. Hence, the required time to classify single fruit image is approximately 0.2 second.