BatNet: a deep learning‐based tool for automated bat species identification from camera trap images
Gabriella Krivek,Alexander Gillert,Martin Harder,Marcus Fritze,Karina Frankowski,Luisa Timm,Liska Meyer‐Olbersleben,Gerald Kerth,Uwe Freiherr von Lukas,Jaap van Schaik
DOI: https://doi.org/10.1002/rse2.339
IF: 5.7874
2023-05-10
Remote Sensing in Ecology and Conservation
Abstract:Infrared light barriers combined with camera traps offer an automated method to monitor bat population dynamics at hibernacula. However, the broad‐scale use of this method is limited by the need to manually identify thousands of images every year. Here, we present a novel, deep learning‐based tool, BatNet, for fast and accurate identification of 13 European bat species or species complexes from camera trap images. The resulting data can be used to accurately describe species‐level ecological metrics, such as species diversity, relative abundance and phenological estimates. BatNet can be retrained to learn new species and specific sites; thus, it vastly expands the scalability and reproducibility of automated bat monitoring and data‐driven bat conservation efforts. Automated monitoring technologies can increase the efficiency of ecological data collection and support data‐driven conservation. Camera traps coupled with infrared light barriers can be used to monitor temperate‐zone bat assemblages at underground hibernacula, where thousands of individuals of multiple species can aggregate in winter. However, the broad‐scale adoption of such photo‐monitoring techniques is limited by the time‐consuming bottleneck of manual image processing. Here, we present BatNet, an open‐source, deep learning‐based tool for automated identification of 13 European bat species from camera trap images. BatNet includes a user‐friendly graphical interface, where it can be retrained to identify new bat species or to create site‐specific models to improve detection accuracy at new sites. Model accuracy was evaluated on images from both trained and untrained sites, and in an ecological context, where community‐ and species‐level metrics (species diversity, relative abundance, and species‐level activity patterns) were compared between human experts and BatNet. At trained sites, model performance was high across all species (F1‐score: 0.98–1). At untrained sites, overall classification accuracy remained high (96.7–98.2%), when camera placement was comparable to the training images (<3 m from the entrance; 3 m or >45° angle), retraining the detector model with 500 site‐specific annotations achieved an accuracy of over 95% at all sites. In the ecological case study, all investigated metrics were nearly identical between human experts and BatNet. Finally, we exemplify the ability to retrain BatNet to identify a new bat species, achieving an F1‐score of 0.99 while maintaining high classification accuracy for all original species. BatNet can be implemented directly to scale up the deployment of camera traps in Europe and enhance bat population monitoring. Moreover, the pretrained model can serve as a baseline for transfer learning to automatize the image‐based identification of bat species worldwide.
ecology,remote sensing