Data augmentation for the classification of North Atlantic right whales upcalls

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.

Performance of a deep neural network at detecting North Atlantic right whale upcalls

Passive acoustics provides a powerful tool for monitoring the endangered North Atlantic right whale (Eubalaena glacialis), but robust detection algorithms are needed to handle diverse and variable acoustic conditions and differences in recording techniques and equipment. This paper investigates the potential of deep neural networks (DNNs) for addressing this need. ResNet, an architecture commonly used for image recognition, was trained to recognize the time-frequency representation of the characteristic North Atlantic right whale upcall. The network was trained on several thousand examples recorded at various locations in the Gulf of St. Lawrence in 2018 and 2019, using different equipment and deployment techniques. Used as a detection algorithm on fifty 30-min recordings from the years 2015-2017 containing over one thousand upcalls, the network achieved recalls up to 80% while maintaining a precision of 90%. Importantly, the performance of the network improved as more variance was introduced into the training dataset, whereas the opposite trend was observed using a conventional linear discriminant analysis approach. This study demonstrates that DNNs can be trained to identify North Atlantic right whale upcalls under diverse and variable conditions with a performance that compares favorably to that of existing algorithms.