Varroa Mite Detection Using Deep Learning Techniques.

The varroa mite is a major problem for beekeeping today because it threatens the survival of hives. This paper develops deep learning methods for detecting varroa in images to monitor the level of infestation of the hives in order to use treatments against varroa in time and save the bees. The ultimate goal is its implementation by beekeepers. Therefore, the deep learning models are trained on pictures taken by smartphone cameras covering the entire board where both pupae and varroas are placed. This makes the object detection task a challenge, since it becomes a small object detection problem. This paper shows the experiments that have been developed to solve this challenge, such as the use of super resolution techniques, as well as the difficulties encountered.