Filipe, L.V., Canelas, J., Vieira, M., Fonseca, F.C., Cid, A., Castro, J., Machado, I. (2026). Advancing marine mammal monitoring: Large-scale UAV delphinidae datasets and robust motion tracking for group size estimation. Machine Learning with Applications, 23. https://doi.org/10.1016/j.mlwa.2025.100808
The fourth paper with the AIMM name to be published this year just came out. We are thrilled to share a new collaborative publication with WavEC Offshore Renewables, where we explore how AI can help monitor dolphins from drone footage.
This study presents an end-to-end automated system for detecting and tracking dolphins in UAV (drone) footage, developed by WavEC Offshore Renewables in collaboration with AIMM. Using YOLO11 for detection and BoT-SORT for multi-object tracking, the authors show that AI can reliably follow individuals across video frames, even under challenging sea states, glare and variable light, reducing manual analysis time and increasing efficiency in monitoring campaigns.
A key contribution is the creation of a large UAV dolphin dataset (64,705 images, 225,305 boxes) including a dedicated tracking set with 603 annotated trajectories.
Why this is important for AIMM’s work:
This technology allows us to analyse large volumes of field data faster and more consistently, supporting long-term monitoring and reducing observer effort. Automated tracking can greatly enhance our capacity to detect trends, quantify group size, and ultimately help improve conservation efforts and projects towards dolphin populations.
We highly appreciate the collaborative efforts of this study!
Consult the open access article here.