Detection of Tipburn Stress on Lettuce Grown in a Plant Factory using Artificial Intelligence (AI) Models

Lettuce grown in plant factories in fully controlled environments is susceptible to tipburn, a physiological disorder. A crucial factor influencing indoor plant growth is the ability to modify the growing environment to accelerate crop growth. However, tipburn is exacerbated by the rapid growth process, especially in lettuce. This paper presents an automated method by which to detect tipburn in lettuce grown in a plant factory using deep-learning-based Python and Orange 3 with the objective of finding the optimal model for classifying images of tipburn and healthy leaves. Tipburn lettuce images were captured after transplantation. In Python, 26 different models were used, whereas the Inception V3 model was applied in Orange 3 to assess the degree of accuracy. In total, 447 images were used for training with multiple neural network models. To train all of the models, 447 class datasets, 50 epochs, 32 batch sizes, the Adam optimizer, and a learning rate of 0.001 were used in the Python neural network models, whereas in Orange 3, the Adam solver, the rectified linear unit function (ReLU) activation function, 100 neurons in hidden layers, a maximum of 200 iterations in the neural network, and 0.0001 regularizations were applied. The highest accuracy of the models was detected when Python was used. In Python, MobileNetV2 exhibited the highest accuracy (0.933), followed by DenseNet169 (0.911), DenseNet121 (0.900), InceptionResNetV2 (0.900), and MobileNet (0.900). However, MobileNet exhibited the shortest training time of 758.36 seconds, whereas MobileNetV2 required 758.36 seconds. In addition, the precision of the Orange 3 Inception V3 model was 0.862, which was less accurate than that of the Python MobileNetV2 model. The F1 and recall scores of Inception V3 for Orange 3 were 0.865 and 0.873, respectively. Therefore, the detection of lettuce tipburn grown in a plant factory can be achieved using MobileNetV2 with reliable overall accuracy. The early detection of tipburn is beneficial for growers to manipulate controlled environmental parameters to improve the freshness of lettuce grown in plant factories.