Deep learning for white cabbage seedling prediction

Abstract In this study, the classification of white cabbage seedling images is modeled with convolutional neural networks. We focus on a dataset that tracks the seedling growth over a period of 14 days, where photos were taken at four specific moments. The dataset contains 13,200 individual seedlings with corresponding labels and was retrieved from Bejo, a company operating in agriculture. Different pre-trained convolutional neural network and multi-layer perceptron architectures are developed, along with a traditional statistical method, logistic regression. The models are trained to predict the (un) successful growth of the seedlings. We find that the convolutional neural networks outperform the other models, where AlexNet is the best performing model in this research. On the test set, AlexNet is able to classify 94% of the seedlings accurately with an area under the curve of 0.95. Accordingly, AlexNet is shown to be useful and robust in this particular classification task. AlexNet can be further deployed as an early warning tool to aid professionals in making important decisions. Additionally, this model can be further developed to automate the process.