Search for: All records

The LabelBee system is a web application designed to facilitate the collection, annotation and analysis of large amounts of honeybee behavior data from video monitoring. It is developed as part of NSF BIGDATA project “Large-scale multi-parameter analysis of honeybee behavior in their natural habitat”, where we analyze continuous video of the entrance of bee colonies. Due to the large volume of data and its complexity, LabelBee provides advanced Artificial Intelligence and visualization capabilities to enable the construction of good quality datasets necessary for the discovery of complex behavior patterns. It integrates several levels of information: raw video, honeybee positions, decoded tags, individual trajectories and behavior events (entrance/exit, presence of pollen, fanning, etc.). This integration enables the combination of manual and automatic processing by the biologist end-users, who also share and correct their annotation through a centralized server. These annotations are used by the Computer Scientists to create new automatic models, and improve the quality of the automatic modules. The data constructed by this semi-automatized approach can then be exported for the analytic part, which is taking place on the same server using Jupyter notebooks for the extraction and exploration of behavior patterns. 

In this paper, the recognition of pollen bearing honey bees from videos of the entrance of the hive is presented. This computer vision task is a key component for the automatic monitoring of honeybees in order to obtain large scale data of their foraging behavior and task specialization. Several approaches are considered for this task, including baseline classifiers, shallow Convolutional Neural Networks, and deeper networks from the literature. The experimental comparison is based on a new dataset of images of honeybees that was manually annotated for the presence of pollen. The proposed approach, based on Convolutional Neural Networks is shown to outperform the other approaches in terms of accuracy. Detailed analysis of the results and the influence of the architectural parameters, such as the impact of dedicated color based data augmentation, provide insights into how to apply the approach to the target application.