Automatic localization and decoding of honeybee markers using deep convolutional neural networks
Tim Landgraf, Benjamin Wild, Leon Sixt – 2018
The honeybee is a fascinating model for the emergence of collective behavior. Understanding how individual (inter-) actions of thousands of individuals that comprise a honeybee colony requires long-term tracking and identification of the animals with high accuracy. This paper proposes a software pipeline based on two deep convolutional neural networks for the localization and decoding of custom binary markers that can be attached to the honeybee's thoraxes. We show that this approach significantly outperforms a pipeline using more conventional computer vision techniques. We hope that the resulting datasets will be a fundamental asset in the understanding of collective intelligence.