Research library

Towards dense object tracking in a 2D honeybee hive

From human crowds to cells in tissue, the detection and efficient tracking of multiple objects in dense configurations is an important and unsolved problem. In the past, limitations of image analysis restricted studies of dense groups to tracking a single or subset of marked individuals, or to coarse-grained group-level dynamics, all of which yield incomplete information. Here, the authors combine convolutional neural networks (CNNs) with the model environment of a honeybee hive to automatically recognize all individuals in a dense group from raw image data. A new, adapted individual labeling scheme is created and the segmentation architecture U-Net is used with a loss function dependent on both object identity and orientation. Temporal regularities of the video recording are exploited in a recurrent manner, reducing network size by 94% compared to the original U-Net while achieving near human-level performance. The dataset contains over 375,000 labeled bee instances across 720 video frames at 2 FPS. The method correctly detects 96% of individuals with a location error of ~7% of a typical body dimension and orientation error of ~12°, and provides an important step toward automated dense object tracking in biological systems.