My co-authors and I are pleased to share our open-access publication: "UAV-Based Dolphin Tracking Enables Automated Quantification of Spatial Group Cohesion in Natural and Exposure Settings"
https://onlinelibrary.wiley.com/doi/10.1111/mms.70268 ABSTRACT Changes in spatial group cohesion have been documented in dolphin groups exposed to tourism activities, signaling an avoidance response. To date, dolphin spatial group cohesion is typically analyzed qualitatively because obtaining fine-scale spatial data for gregarious dolphin species is challenging. Unoccupied aerial vehicles (UAV) can record videos from which inter-individual distances (IIDs) can be quantified. Current methods involve time-consuming manual annotation of each individual dolphin in a frame. Here, we present a marine animal-spatial group cohesion pipeline that (1) standardizes the UAV flight protocol, (2) automatically detects and distinguishes dolphins from other objects, (3) calculates geographic positions of each dolphin, adjusted for changes in UAV yaw and camera gimbal pitch, and (4) quantifies and visualizes IIDs under different environmental conditions allowing comparison of IIDs before, during, and after exposure to anthropogenic activities. To test the application of the pipeline, we exposed common dolphins (Delphinus delphis) to experimental vessel approaches. In total, 2572 dolphins were manually annotated, reaching 13,445 dolphin detections through corrective annotation. Using the final detection model 2,698,099 dolphins were localized in 156,210 frames of 227 videos at 1 fps. We achieved a dolphin detection precision of 0.97 and recall of 0.91. Our pipeline enables dolphin IID calculations from video frames containing hundreds of dolphins in a time-efficient way. The pipeline is suitable for a range of different UAV-based wildlife studies and can be used to further develop the extraction of other individual dolphin behavior or other spatial group cohesion parameters. If you have any questions you can contact me directly at [email protected]. Best regards, Fadia Al Abbar
_______________________________________________ MARMAM mailing list [email protected] https://lists.uvic.ca/mailman/listinfo/marmam
