Passive acoustic monitoring (PAM) is a useful technique for monitoring marine mammals. However, the quantity of data collected through PAM systems makes automated algorithms for detecting and classifying sounds essential. Deep learning algorithms have shown great promise in recent years, but their performance is limited by the lack of sufficient amounts of annotated data for training the algorithms. This work investigates the benefit of augmenting training datasets with synthetically generated samples when training a deep neural network for the classification of North Atlantic right whale (Eubalaena glacialis) upcalls. We apply two recently proposed augmentation techniques, SpecAugment and Mixup, and show that they improve the performance of our model considerably. The precision is increased from 86% to 90%, while the recall is increased from 88% to 93%. Finally, we demonstrate that these two methods yield a significant improvement in performance in a scenario of data scarcity, where few training samples are available. This demonstrates that data augmentation can reduce the annotation effort required to achieve a desirable performance threshold.

Passive acoustic monitoring is a useful technique for studying aquatic animals, but sustained observing systems require automated algorithms for detecting and classifying sounds of interest. In the last decade, deep neural networks have proven highly successful at solving a wide range of pattern recognition tasks, and recently, we have seen the first promising applications of deep neural networks to detection and classification tasks in marine bioacoustics. Deep neural networks exhibit a high degree of versatility and adaptability: the same network architecture can be trained to accomplish a multitude of tasks by feeding appropriate training data to the network without the need to modify the underlying algorithm. Thus, neural networks have the potential to transform our approach to developing acoustic detection and classification programs, enabling researchers in the field to develop or re-purpose their own programs. MERIDIAN is contributing towards this goal through the development of the open-source Python package Ketos, which provides a high-level programming interface for building training datasets and developing neural network based detectors and classifiers for analyzing underwater acoustics data. In this contribution, an overview of the software package will be given and its functionalities will be demonstrated through case studies.

about 30 researchers gathered in Victoria, BC, for the workshop Detection and Classification in Marine Bioacoustics with Deep Learning organized by MERIDIAN and hosted by Ocean Networks Canada. The workshop was attended by marine biologists, data scientists, and computer scientists coming from both Canadian coasts and the US and representing a wide spectrum of research organizations including universities, government (Fisheries and Oceans Canada, National Oceanic and Atmospheric Administration), industry (JASCO Applied Sciences, Google, Axiom Data Science), and non-for-profits (Orcasound, OrcaLab). Consisting of a mix of oral presentations, open discussion sessions, and hands-on tutorials, the workshop program offered a rare opportunity for specialists from distinctly different domains to engage in conversation about Deep Learning and its promising potential for the development of detection and classification algorithms in underwater acoustics. Presentations given at the workshop can be found on MERIDIAN's website at meridian.cs.dal.ca, while the hands-on tutorial can be found at gitlab.meridian.cs.dal.ca/workshops/victoria nov2019. In the following, we summarize key points from the presentations and discussion sessions. The list of participants can be found at the end of the report.

No abstract

Bird population census is an important indicator in conservation programs. However, the process of detecting and identifying particular species is time-consuming and challenging, often being conducted in remote locations. In this scenario, the development of automated acoustic systems for bird monitoring is crucial. In this study, we propose a simple but effective 3-step approach for identifying the Amazona rhodocorytha, an endangered Brazilian parrot, among 4 other species belonging to the same family. This approach consists of a pre-processing step, a feature extraction step using the MFCC algorithm and a classification step by employing an Artificial Neural Network. Results show that the proposed approach is both suitable and robust for this type of application, achieving excellent classification results of up to 98% accuracy.