Bio-acoustics & Artificial Intelligence
Whistles characterisation using artificial intelligence reveals responses of short-beaked common dolphins to a bio-inspired acoustic mitigation device for fishing nets
This study introduces a novel semi-automated deep learning method, DYOC, for efficiently extracting whistle contours from acoustic recordings of short-beaked common dolphins. The application of DYOC to a large dataset from the Bay of Biscay reveals complex dolphin acoustic behaviors influenced by fishing nets and a bio-inspired acoustic beacon. This AI-driven approach significantly accelerates annotation and provides critical insights for conservation efforts.
Executive Impact & Key Findings
Our analysis reveals the most significant quantifiable impacts this AI innovation can bring to your enterprise operations:
6x
Faster Annotation Speed
8,730
Whistles Annotated
69.2%
YOLO mAP50 Score
6.1%
ResNet18 RMSE
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Revolutionizing Bioacoustic Data Annotation
Our novel DYOC method offers a semi-automated approach to whistle contour extraction, significantly accelerating the annotation process while maintaining high accuracy, crucial for large-scale bioacoustic research.
Enterprise Process Flow
Raw Audio Ingestion
→
Spectrogram Conversion
→
YOLOv8 Bbox Prediction
→
ResNet18 Contour Regression
→
Manual Correction & Refinement
The integration of YOLOv8 for object detection and ResNet18 for precise contour regression, combined with strategic manual verification, sets a new standard for efficiency in bioacoustic data processing. This hybrid approach capitalizes on AI's speed for initial predictions and human expertise for complex scenarios.
Unveiling Complex Acoustic Behaviors
DYOC enables a detailed analysis of short-beaked common dolphin whistles, revealing nuanced responses to environmental stimuli such as fishing nets and an acoustic mitigation beacon. These insights are vital for understanding dolphin communication and behavioral ecology.
17.95 dB
Lower SNR with Fishing Net Present, indicating avoidance.
When fishing nets are present, dolphins exhibit significantly lower signal-to-noise ratios in their whistles, indicating either increased distance or altered vocalization intensity. This suggests a behavioral response to avoid the perceived threat.
| Behavioral Context | Key Whistle Characteristics | Implication |
|---|---|---|
| Milling/Resting |
|
Suggests continued information exchange, possibly for group cohesion. |
| Beacon Activation |
|
Distinct acoustic response to beacon introduction, potentially for communication. |
| Foraging vs. Travelling (Net Present) |
|
Modulated whistling in response to fishing net based on current activity. |
These findings highlight the dynamic and context-dependent nature of dolphin whistle production, providing crucial data for effective management and conservation strategies.
Informing Conservation and Mitigation Strategies
The detailed whistle characterization facilitated by AI offers unprecedented insights into the effectiveness of acoustic mitigation devices and the impact of anthropogenic activities on marine mammal populations, directly informing conservation policies.
Case Study: DOLPHINFREE Project - Bay of Biscay
The DOLPHINFREE project developed a bio-inspired acoustic beacon to deter dolphins from fishing nets, addressing critical bycatch issues. Our AI analysis directly supported the evaluation of this device's effectiveness.
Our study observed significant variations in dolphin whistle characteristics—including duration, frequency, and complexity (number of inflections)—in response to the beacon's activation and the presence of fishing nets. These behavioral shifts suggest that dolphins are actively reacting to the mitigation device and environmental stressors.
Key Outcome: The findings inform an action plan for 2024-2026 to assess and optimize the DOLPHINFREE beacon, aiming to reduce bycatch of short-beaked common dolphins in the Bay of Biscay.
By understanding how dolphins adapt their communication in challenging environments, we can design more effective acoustic deterrents and inform policy decisions that better protect these vulnerable populations.
Advanced ROI Calculator
Estimate the potential financial impact of this AI solution for your organization.
Annual Savings Potential
Annual Hours Reclaimed
Your Implementation Roadmap
A phased approach to integrate this AI solution seamlessly into your existing operations:
Phase 1: Pilot & Data Integration (Weeks 1-4)
Initial setup of DYOC, integration with existing acoustic data archives, and a pilot study on a small subset of recordings to fine-tune AI models for specific datasets and whistle characteristics. Establish baseline performance and gather feedback.
Phase 2: Full-Scale Deployment & Training (Months 2-3)
Expand DYOC application to full datasets, including real-time or near real-time acoustic streams. Provide comprehensive training for bioacousticians and research teams on using the semi-automated annotation tool and interpreting AI outputs. Initiate initial large-scale whistle characterization studies.
Phase 3: Advanced Analysis & Integration (Months 4-6)
Utilize the accelerated annotation for deeper behavioral ecology studies, conservation impact assessments, and development of new research questions. Integrate AI-derived insights into existing ecological models and policy recommendations. Explore further model refinements and custom feature extraction for specialized research needs.
Ready to Transform Your Bioacoustic Research?
Unlock unprecedented efficiency and insights in your marine mammal studies. Let's discuss how our AI solutions can accelerate your team's progress and deepen your understanding of marine life.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat