Real-Time Dolphin Whistle Detection on Raspberry Pi Zero 2 W with a TFLite Convolutional Neural Network
The escalating conflict between cetaceans and fisheries underscores the need for efficient mitigation strategies that balance conservation priorities with economic viability. This study presents a TinyML-driven approach deploying an optimized Convolutional Neural Network (CNN) on a Raspberry Pi Zero...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-05-01
|
| Series: | Robotics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2218-6581/14/5/67 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850256412886695936 |
|---|---|
| author | Rocco De Marco Francesco Di Nardo Alessandro Rongoni Laura Screpanti David Scaradozzi |
| author_facet | Rocco De Marco Francesco Di Nardo Alessandro Rongoni Laura Screpanti David Scaradozzi |
| author_sort | Rocco De Marco |
| collection | DOAJ |
| description | The escalating conflict between cetaceans and fisheries underscores the need for efficient mitigation strategies that balance conservation priorities with economic viability. This study presents a TinyML-driven approach deploying an optimized Convolutional Neural Network (CNN) on a Raspberry Pi Zero 2 W for real-time detection of bottlenose dolphin whistles, leveraging spectrogram analysis to address acoustic monitoring challenges. Specifically, a CNN model previously developed for classifying dolphins’ vocalizations and originally implemented with TensorFlow was converted to TensorFlow Lite (TFLite) with architectural optimizations, reducing the model size by 76%. Both TensorFlow and TFLite models were trained on 22 h of underwater recordings taken in controlled environments and processed into 0.8 s spectrogram segments (300 × 150 pixels). Despite reducing model size, TFLite models maintained the same accuracy as the original TensorFlow model (87.8% vs. 87.0%). Throughput and latency were evaluated by varying the thread allocation (1–8 threads), revealing the best performance at 4 threads (quad-core alignment), achieving an inference latency of 120 ms and sustained throughput of 8 spectrograms/second. The system demonstrated robustness in 120 h of continuous stress tests without failure, underscoring its reliability in marine environments. This work achieved a critical balance between computational efficiency and detection fidelity (F1-score: 86.9%) by leveraging quantized, multithreaded inference. These advancements enable low-cost devices for real-time cetacean presence detection, offering transformative potential for bycatch reduction and adaptive deterrence systems. This study bridges artificial intelligence innovation with ecological stewardship, providing a scalable framework for deploying machine learning in resource-constrained settings while addressing urgent conservation challenges. |
| format | Article |
| id | doaj-art-11515f4ebc2d42e78f564b05135ef1d4 |
| institution | OA Journals |
| issn | 2218-6581 |
| language | English |
| publishDate | 2025-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Robotics |
| spelling | doaj-art-11515f4ebc2d42e78f564b05135ef1d42025-08-20T01:56:39ZengMDPI AGRobotics2218-65812025-05-011456710.3390/robotics14050067Real-Time Dolphin Whistle Detection on Raspberry Pi Zero 2 W with a TFLite Convolutional Neural NetworkRocco De Marco0Francesco Di Nardo1Alessandro Rongoni2Laura Screpanti3David Scaradozzi4Institute of Biological Resources and Marine Biotechnology (IRBIM), National Research Council (CNR), 60125 Ancona, ItalyDipartimento di Ingegneria Dell’Informazione, Università Politecnica delle Marche, 60131 Ancona, ItalyDipartimento di Ingegneria Dell’Informazione, Università Politecnica delle Marche, 60131 Ancona, ItalyDipartimento di Ingegneria Dell’Informazione, Università Politecnica delle Marche, 60131 Ancona, ItalyDipartimento di Ingegneria Dell’Informazione, Università Politecnica delle Marche, 60131 Ancona, ItalyThe escalating conflict between cetaceans and fisheries underscores the need for efficient mitigation strategies that balance conservation priorities with economic viability. This study presents a TinyML-driven approach deploying an optimized Convolutional Neural Network (CNN) on a Raspberry Pi Zero 2 W for real-time detection of bottlenose dolphin whistles, leveraging spectrogram analysis to address acoustic monitoring challenges. Specifically, a CNN model previously developed for classifying dolphins’ vocalizations and originally implemented with TensorFlow was converted to TensorFlow Lite (TFLite) with architectural optimizations, reducing the model size by 76%. Both TensorFlow and TFLite models were trained on 22 h of underwater recordings taken in controlled environments and processed into 0.8 s spectrogram segments (300 × 150 pixels). Despite reducing model size, TFLite models maintained the same accuracy as the original TensorFlow model (87.8% vs. 87.0%). Throughput and latency were evaluated by varying the thread allocation (1–8 threads), revealing the best performance at 4 threads (quad-core alignment), achieving an inference latency of 120 ms and sustained throughput of 8 spectrograms/second. The system demonstrated robustness in 120 h of continuous stress tests without failure, underscoring its reliability in marine environments. This work achieved a critical balance between computational efficiency and detection fidelity (F1-score: 86.9%) by leveraging quantized, multithreaded inference. These advancements enable low-cost devices for real-time cetacean presence detection, offering transformative potential for bycatch reduction and adaptive deterrence systems. This study bridges artificial intelligence innovation with ecological stewardship, providing a scalable framework for deploying machine learning in resource-constrained settings while addressing urgent conservation challenges.https://www.mdpi.com/2218-6581/14/5/67TinyMLdolphin whistle detectionconvolutional neural network (CNN)TensorFlow LiteRaspberry Pi |
| spellingShingle | Rocco De Marco Francesco Di Nardo Alessandro Rongoni Laura Screpanti David Scaradozzi Real-Time Dolphin Whistle Detection on Raspberry Pi Zero 2 W with a TFLite Convolutional Neural Network Robotics TinyML dolphin whistle detection convolutional neural network (CNN) TensorFlow Lite Raspberry Pi |
| title | Real-Time Dolphin Whistle Detection on Raspberry Pi Zero 2 W with a TFLite Convolutional Neural Network |
| title_full | Real-Time Dolphin Whistle Detection on Raspberry Pi Zero 2 W with a TFLite Convolutional Neural Network |
| title_fullStr | Real-Time Dolphin Whistle Detection on Raspberry Pi Zero 2 W with a TFLite Convolutional Neural Network |
| title_full_unstemmed | Real-Time Dolphin Whistle Detection on Raspberry Pi Zero 2 W with a TFLite Convolutional Neural Network |
| title_short | Real-Time Dolphin Whistle Detection on Raspberry Pi Zero 2 W with a TFLite Convolutional Neural Network |
| title_sort | real time dolphin whistle detection on raspberry pi zero 2 w with a tflite convolutional neural network |
| topic | TinyML dolphin whistle detection convolutional neural network (CNN) TensorFlow Lite Raspberry Pi |
| url | https://www.mdpi.com/2218-6581/14/5/67 |
| work_keys_str_mv | AT roccodemarco realtimedolphinwhistledetectiononraspberrypizero2wwithatfliteconvolutionalneuralnetwork AT francescodinardo realtimedolphinwhistledetectiononraspberrypizero2wwithatfliteconvolutionalneuralnetwork AT alessandrorongoni realtimedolphinwhistledetectiononraspberrypizero2wwithatfliteconvolutionalneuralnetwork AT laurascrepanti realtimedolphinwhistledetectiononraspberrypizero2wwithatfliteconvolutionalneuralnetwork AT davidscaradozzi realtimedolphinwhistledetectiononraspberrypizero2wwithatfliteconvolutionalneuralnetwork |