Abstract Acoustic cameras, or imaging sonars, are often used to monitor marine energy sites in regions where the water is too dark or turbid for optical sensing. To do so more effectively, scientists are investigating automated detection methodologies to use on these data. However, prior work has found that existing automated detection approaches struggle with the dynamic image background around marine energy devices—such as moving turbine blades. While open-access datasets, methods, and standard evaluation metrics are needed to quickly develop and compare novel automated detection methods, none yet exist for this domain. Using previously collected data, in this work we created a labeled dataset of possible marine life interactions in acoustic camera video around an operating tidal turbine. We call this dataset the Pacific Northwest National Laboratory dataset for Tracking Underwater Nautical Activity around Marine Energy LocaTions or PNNL TUNAMELT dataset. In addition to this dataset, we developed an automated detection pipeline which filters noise from the acoustic camera imagery and then performs object detection to identify possible targets. To analyze our automated detection pipeline, we used a series of common detection and classification metrics. In doing so, we found that our pipeline detected 98% of targets and removed 70% of target-less frames in our dataset. These results illustrate our method’s potential utility as an aid to a human analyst tasked with extracting targets of interest from the dataset. Finally, we openly release our labeled dataset and all associated code to support and encourage future work in this domain.
@article{nowak_lom_2026,author={Nowak, Theodore and Staines, Garrett and Abdullai, Blerim},title={PNNL-TUNAMELT: Toward automating the detection of interactions with marine energy devices using acoustic camera sensors},journal={Limnology and Oceanography: Methods},year={2026},volume={n/a},number={n/a},pages={e70024},doi={https://doi.org/10.1002/lom3.70024},url={https://aslopubs.onlinelibrary.wiley.com/doi/abs/10.1002/lom3.70024},eprint={https://aslopubs.onlinelibrary.wiley.com/doi/pdf/10.1002/lom3.70024},}
2025
Standards
IEEE P7003 Standard for Algorithmic Bias Considerations
@article{ieee-p7003,title={IEEE P7003 Standard for Algorithmic Bias Considerations},author={Weger, Gerlinde and al, et},journal={IEEE Standards Association},year={2025},month=jan,day={24},doi={10.1109/IEEESTD.2025.10851955}}
2024
Conference
Enabling Autonomous Management of Drifting Underwater Floats with an ASV
Trevor W. Harrison, Aaron Marburg, Mitchell Scott, and 5 more authors
@inproceedings{10754147,author={Harrison, Trevor W. and Marburg, Aaron and Scott, Mitchell and Maheshwari, Deven and Noe, Jessica and Pickett, Madison and Cavagnaro, Robert J. and Nowak, Theodore},booktitle={OCEANS 2024 - Halifax},title={Enabling Autonomous Management of Drifting Underwater Floats with an ASV},year={2024},volume={},number={},pages={1-8},keywords={Meters;Sea surface;Target tracking;Mechatronics;Navigation;Robot vision systems;Sea measurements;Sampling methods;Surface treatment;Robots;floats;ASV;autonomy;robotic sampling},doi={10.1109/OCEANS55160.2024.10754147},}
2023
Pre-Print
A Theoretical Perspective on Subnetwork Contributions to Adversarial Robustness
Jovon Craig, Josh Andle, Theodore S. Nowak, and 1 more author
@article{advpose,title={AdvPose: Generating Realistic Adversarial Scenes Through Object Pose Manipulation},author={Nowak, Theodore S. and Slyman, Eric B.},journal={Sponsor Report},year={2022}}
2020
Controlled
Adversarial Attacks Against LiDAR: Exploring the Vulnerabilities of LiDAR and Point Cloud-based Deep Learning Models
@article{adv-lidar,title={Adversarial Attacks Against LiDAR: Exploring the Vulnerabilities of LiDAR and Point Cloud-based Deep Learning Models},author={Nowak, Theodore S.},journal={Sponsor Report},year={2020}}
2018
Pre-Print
Deep Net Triage: Assessing the Criticality of Network Layers by Structural Compression
@article{deep-net-triage,author={Nowak, Theodore S. and Corso, Jason J.},title={Deep Net Triage: Assessing the Criticality of Network Layers by Structural Compression},journal={Arxiv},volume={abs/1801.04651},year={2018},archiveprefix={arXiv},}
2015
Journal
Seizure reduction through interneuron-mediated entrainment using low frequency optical stimulation
Thomas P. Ladas, Chia-Chu Chiang, Luis E. Gonzalez-Reyes, and 2 more authors
@article{ladas-seizure,title={Seizure reduction through interneuron-mediated entrainment using low frequency optical stimulation},author={Ladas, Thomas P. and Chiang, Chia-Chu and Gonzalez-Reyes, Luis E. and Nowak, Theodore S. and Durand, Dominique M.},journal={Experimental neurology},volume={269},pages={120--132},year={2015},publisher={Elsevier},}