StrateSea Technology has demonstrated the capabilities of its operator-driven platform, FeatureLab, during an independent evaluation conducted by an EU Navy in collaboration with a European maritime defense industry partner.
The field test evaluated FeatureLab in a mine countermeasures post-mission analysis (PMA) workflow, comparing automated target recognition (ATR) against conventional operator-driven analysis using established PMA software. No StrateSea personnel were on site, and no in-person operator training was conducted.
Rapid Analysis & Superior Target Detection
The evaluation demonstrated significant efficiency gains and improved detection accuracy when utilizing automated processes:
FeatureLab ATR analysis was completed in approximately 3 minutes, compared with roughly 29 to 36 minutes for participating human operators.
The FeatureLab ATR model detected 92% of the reference targets, compared to approximately 78% to 85% detected by human operators.
Manual results varied with operator experience, particularly in false-contact generation. FeatureLab provided a repeatable ATR analytical baseline while preserving human judgment for final decisions.
Mission-Tunable Performance & Rapid Adaptation
The platform proved capable of rapid deployment and iterative refinement using minimal initial data. Initial FeatureLab models were developed from just 64 labeled examples for one target class and 47 for another, totaling 111 examples, rather than relying on a large mature training data set. Adding just 17 new contacts and retraining mid-mission materially improved all performance metrics.
Operators could adjust model confidence thresholds to balance sensitivity against false-alert burden. Increasing the confidence threshold substantially reduced false detections, allowing the operating point of the ATR system to be tuned to mission priorities rather than remaining fixed by the algorithm.
Local operators were able to build and tune mission-relevant ATR from limited data while reducing analysis time by approximately 90% and maintaining strong target-detection performance. FeatureLab demonstrated that ATR can adapt at the speed of the mission, not the speed of a traditional development cycle.
