BeeHealthClassifier

Convolutional Neural Network (CNN) model to classify bees by health condition.

Bees are essential for the stability of the environ-ment, as they enable the production of seeds and fruits. One-third of all our food depends on the pollination process. Withoutpollinators, food production would be insufficient for the worldpopulation. Unfortunately, beehives are susceptible to severaldiseases, some very contagious, such as varroa mite, a naturalpredator. Recognizing the first signs of disease in the hivesis essential. The earlier a disease is detected, the less likely itis that a hive will collapse. The traditional way of obtaininginformation about the hives health is to review it directly insideit. However, this process hinders the workflow of bees. By analyzing images of bees coming out of the hive, we can gain a greater understanding of them. For example, a hive infected withvarroa mites will have bees with deformed wings or mites on theirbacks. Deep Learning (DL), in particular Convolutional NeuralNetworks (CNNs), have demonstrated outstanding performanceto fields such as computer vision, speech recognition amongothers, where they have produced results comparable to and insome cases surpassing human expert performance. In this work, we propose building CNN-based solutions trained to classify bees by their health condition by using as input the bee image.