Bumblebee Re-Identification Dataset

Frederic Tausch, Simon Stock, Julian Fricke, Olaf Klein; Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Workshops, 2020, pp. 35-37

Over the past decade, entomologists have observed a worldwide decline in the population of pollinating insects. Even though there is a number of hypothesesabout of potential underlying causes for this decline, thereare concerns that insecticides are at least partially responsible. The design of insecticides requires accuraterisk assessment procedures to avoid damage to beneficial in-sects and like pollinators such as the buff-tailed bumblebee(Bombus terrestris). Therefore, current guideline drafts by the EFSA include homing fight studies to test insect behavioral performance. Since these studies (e.g., homing study) inspect the individual's change in behavior, insect re-ID (reidentification) is crucial. Current bumblebee re-ID techniques are limited to placing markers on individual subjects, which can be analog or digital (RFID). These steps are very time consuming, intrusive and are prone to mechanical failure. To find a solution to demedy these shortcomings, a visual approach is pro-posed. Therefore we propose the following bumblebee re-ID dataset in order to build such systems.