Figure 3 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Source publication
Artificial intelligence (AI) has become a significantly growing field in the environmental sector due to its ability to solve problems, make decisions, and recognize patterns. The significance of AI in wildlife acoustic monitoring is particularly important because of the vast amounts of data that are available in this field, which can be leveraged...
Context in source publication
Context 1
... technology has been applied in various aspects of acoustic wildlife monitoring including behavioral patterns [22,23,26], species recognition [24,25,29], species density/population estimation, species diversity [30], and illegal trade [32] (Table 1 and Figure 3). These applications have helped researchers understand the location and migration patterns of animals, as well as their social group affiliations. ...
Similar publications
Ecoacoustic monitoring has proliferated as autonomous recording units (ARU) have become more accessible. ARUs provide a non-invasive, passive method to assess ecosystem dynamics related to vocalizing animal behavior and human activity. With the ever-increasing volume of acoustic data, the field has grappled with summarizing ecologically meaningful...
The detection and localization of acoustic sources remain technological challenges in bioacoustics, in particular, the tracking of moving underwater sound sources with a portable waterproof tool. For instance, this type of tool is important to describe the behavior of cetaceans within social groups. To contribute to this issue, an original innovati...
Citations
... Passive acoustic monitoring represents an innovative tool for recording both soundproducing species and environments (Do Nascimento et al., 2020; S. Sharma et al., 2023;Sugai et al., 2019;Usman et al., 2020). This tool allows researchers to survey species over large spatial-temporal scales while minimising the fieldwork efforts (Hoyos-Cardona et al., 2021). ...
The study of nocturnal animal populations has long been neglected, leading to what is known as the "nocturnal problem." However, advancements in technology and analytical techniques have opened new possibilities, with Passive Acoustic Monitoring (PAM) emerging as a promising approach. PAM data has become a valuable input for ecological models, including occupancy models, enabling researchers to explore factors influencing species' habitat use. In the biodiverse Cauca River Canyon, Colombia, we investigated species-specific responses of nocturnal birds to environmental variables using PAM techniques and ARBIMON's semi-automatic annotation workflow. We were able to model the occupancy of Megascops choliba, Nyctidromus albicollis and Pulsatrix perspicillata and discover that the key variables for each species changed as the regional seasons changed. Our findings highlight the suitability of data acquired through PAM and supported by detection algorithms in modeling the occupancy of nocturnal bird species and its potential in addressing the "nocturnal problem”. Despite progress, further developments are needed to fully harness the power of PAM in understanding ecological dynamics and conserving nocturnal ecosystems. This study contributes to the advancement of methodologies for studying nocturnal bird populations and underscores the importance of considering habitat characteristics and seasonality to further understanding the ecology of these elusive species.
... Traditional methods, such as manual surveys and visual inspections, rely heavily on human expertise and can be prone to errors, particularly in challenging environments where species are difficult to detect (Beijbom et al., 2015). In contrast, AI-based methods can process and analyze vast amounts of data quickly and accurately, enabling the detection of invasive species in real-time and across large areas (Sharma et al., 2023). ...
... Additionally, collection of information using ARUs can result in large datasets, which can be incredibly time consuming and labor intensive to clean, code, and analyze to extract meaningful trends (Priyadarshani et al. 2018). New technologies, such as the development of Artificial Intelligence (AI), are able to overcome some of these limitations by assisting in data cleaning and analysis (Sharma et al. 2023), but trained human observers are still more proficient at identifying misclassifications in the datasets. ...
This study explores the increasing use of autonomous recording units (ARUs) in wildlife surveys. While ARUs offer cost-effective and efficient data collection, challenges arise in analyzing large datasets and accurately assessing species abundance. Our research focuses on avian communities, emphasizing the impact of vocal mimicry by Northern Mockingbirds ( Mimus polyglottos ) on survey accuracy. Utilizing the Merlin Bird ID application, we found an average accuracy rate of ~81.3%, with mockingbirds contributing ~31% of false positive identifications. Finding potential solutions for distinguishing mimics in bioacoustic survey data is crucial for enhancing accuracy as researchers increasingly adopt this methodology in the future.
... Environmental monitoring relies heavily on AI-driven technologies like remote sensing and machine learning (Garcia, 2023;Miller et al., 2023;Neo et al., 2023;Sharma et al., 2023;Singh, 2023). Using AI algorithms on satellite data, these technologies can accurately identify deforestation, assess water quality, and monitor wildlife populations. ...
Artificial intelligence (AI) is an umbrella term for a wide range of machine intelligence systems that can replicate the behavior of humans. AI and Big Data have emerged as defining characteristics of the fourth industrial revolution (IR). AI has developed tools. Because of the novelty, the investigation of IR, AI, and their environmental effects is still in the early stages of exploration. This study investigates how IR and AI affect human and environmental health and also discusses IR, AI, machine-human ideas, innovation, and AI's environmental benefits, further examines the challenges of these innovations, and recommends additional studies to explain their progress. As a result, the application of AI technology in environmental management, particularly concerning pollution, has become a significant advancement in reshaping our approach to monitoring the environment. Numerous countries are reaping substantial advantages by integrating AI in creating, executing, and assessing measures to address environmental degradation. These innovations can yield societal advantages and contribute to achieving the Sustainable Development Goals (SDGs) 2030; unfortunately, it is important to acknowledge that these benefits may not align well with environmental sustainability objectives, and the increasing number of electronic gadgets presents an additional concern. Conducting future research is crucial to investigate the growing prevalence of electronic devices utilized for AI, its potential ramifications for the future trajectory of climate change, and the approaches being taken to address the issue. Future research should prioritize conducting lifecycle environmental impact analyses, developing sustainable AI hardware, optimizing renewable energy usage, advancing climate modeling techniques, finding effective solutions for managing e-waste, utilizing AI for environmental monitoring and protection, conducting socio-environmental impact studies, developing policies and regulations, creating energy-efficient AI algorithms, and integrating circular economy principles to ensure that AI advancements align with environmental sustainability.
... We conducted a thorough analysis of our model's performance using various matrices. The confusion matrices provided insight into true positives (TP), which are instances where the model's predicted and observed outcomes were both true; true negatives, (TN), which are instances where the model correctly predicted and observed false outcomes; false positives (FP), which are instances where the model incorrectly predicted true outcomes; and false negatives (FN), which are instances where the model incorrectly predicted false outcomes [42]. These metrics offer a comprehensive assessment of multiclass classification effectiveness, accommodating datasets with both even and uneven distributions [7]. ...
... Our evaluation encompassed several key metrics. Accuracy [42], a widely utilized measure, is the ratio of correct predictions to the total number of predictions determined by ...
Accurately identifying individual wildlife is critical to effective species management and conservation efforts. However, it becomes particularly challenging when distinctive features, such as spot shape and size, serve as primary discriminators, as in the case of Sika deer. To address this challenge, we employed four different Convolutional Neural Network (CNN) base models (EfficientNetB7, VGG19, ResNet152, Inception_v3) within a Siamese Network Architecture that used triplet loss functions for the identification and re-identification of Sika deer. Subsequently, we then determined the best-performing model based on its ability to capture discriminative features. From this model, we extracted embeddings representing the learned features. We then applied a Support Vector Machine (SVM) to these embeddings to classify individual Sika deer. We analyzed 5169 image datasets consisting of images of seven individual Sika deers captured with three camera traps deployed on farmland in Hokkaido, Japan, for over 60 days. During our analysis, ResNet152 performed exceptionally well, achieving a training accuracy of 0.97, and a validation accuracy of 0.96, with mAP scores for the training and validation datasets of 0.97 and 0.96, respectively. We extracted 128 dimensional embeddings of ResNet152 and performed Principal Component Analysis (PCA) for dimensionality reduction. PCA1 and PCA2, which together accounted for over 80% of the variance collectively, were selected for subsequent SVM analysis. Utilizing the Radial Basis Function (RBF) kernel, which yielded a cross-validation score of 0.96, proved to be most suitable for our research. Hyperparameter optimization using the GridSearchCV library resulted in a gamma value of 10 and C value of 0.001. The OneVsRest SVM classifier achieved an impressive overall accuracy of 0.97 and 0.96, respectively, for the training and validation datasets. This study presents a precise model for identifying individual Sika deer using images and video frames, which can be replicated for other species with unique patterns, thereby assisting conservationists and researchers in effectively monitoring and protecting the species.
... Research has shown that deep learning algorithms work well for bird audio classification [18,19]. Classifiers that use deep learning can handle large amounts of data [20]. Deep learning algorithms such as CNNs are good at recognizing patterns and are often used to analyse sounds and find patterns [21]. ...
To recognize birds based on their calls, it would be helpful to have access to a machine-learning system. Researchers use machine learning and artificial intelligence (AI) algorithms to identify and differentiate bird calls. In this respect, convolutional neural networks (CNNs) are robust machine learning toolkits that have shown success in the field of sound. However, these AI and machine learning algorithms are not intelligible and cannot be interpreted. Therefore, it is challenging to comprehend how these algorithms conclude that birds may be identified based on their calls. These algorithms are sometimes called “black boxes” for these reasons. This study aims to develop both explainable and interpretable techniques to categorize birds based on their sounds. With a focus on the interpretability of features by the convolutional filters and how these characteristics contribute to classification, we empirically evaluate two well-known explainer/interpretable methodologies called LIME (local interpretable model-agnostic explanations) and SHAP (SHAPley additive explanations) to determine the interpretability of our proposed model, which is used for the categorization of species from their sound. Our model achieves 92% accuracy while being simpler and having fewer layers than competing models. Because of eXplainable AI (XAI), the model is not only better but also more reliable. To our knowledge, this is the first time that XAI has been used for the purpose of identifying bird calls. The results showed that SHAP performed slightly better than LIME regarding identity, stability, and separability.
... Additional approaches employ Empirical Mode Decomposition techniques [30]. The extensive volume of passive acoustic data collected over past decades, coupled with technological advancements in deep learning, has stimulated significant interest within the artificial intelligence community towards applications in animal communication [31]. In the context of data sharing and standardization: in this context, BEANS [11] represented a substantial contribution by introducing an integrative framework. ...
Marine mammal communication is a complex field, hindered by the diversity of vocalizations and environmental factors. Accurate classification of these vocalizations is critical for understanding species behavior, monitoring population trends, and assessing the impact of human activities on marine life. However, current classification approaches face significant challenges due to the wide range of vocalization types and environmental noise. The Watkins Marine Mammal Sound Database (WMMD) constitutes a comprehensive labeled dataset employed in machine learning applications. Nevertheless, the methodologies for data preparation, preprocessing, and classification documented in the literature exhibit considerable variability and are typically not applied to the dataset in its entirety. This study initially undertakes a concise review of the state-of-the-art benchmarks pertaining to the dataset, with a particular focus on clarifying data preparation and preprocessing techniques. Subsequently, we explore the utilization of the Wavelet Scattering Transform (WST) and Mel spectrogram as preprocessing mechanisms for feature extraction. In this paper, we introduce WhaleNet (Wavelet Highly Adaptive Learning Ensemble Network), a sophisticated deep ensemble architecture for the classification of marine mammal vocalizations, leveraging both WST and Mel spectrogram for enhanced feature discrimination. By addressing the inconsistencies in data preparation and by utilizing advanced preprocessing techniques, our approach provides a more robust framework for classifying marine mammal vocalizations, which is essential for conservation efforts and behavioral research. By integrating the insights derived from WST and Mel representations, we achieved an improvement in classification accuracy by 8−10% over existing architectures, corresponding to a classification accuracy of 97.61%.
... The monitoring of wildlife is essential for wildlife conservation and biodiversity management. While camera traps were the tool of choice for monitoring many species [1,2], the practicality of relying solely on visual data diminished for certain species because of factors such as body size and behavior [3,4]. Passive acoustic monitoring (PAM) emerged as an alternative to collect large datasets [5], especially for avian species. ...
... With an "image" as input, detection and classification models could be applied to the converted audio recording. But the lack of comprehensive labeled datasets caused, that existing models often exhibit species specificity, limited geographical cover-age or required re-training and programming expertise to be applied to new case studies [14,9,4]. To the best of our knowledge, available ready-to-use models have not been developed to process large datasets or they required a certain amount of manual labeling. ...
Analyses for biodiversity monitoring based on passive acoustic monitoring (PAM) recordings is time-consuming and challenged by the presence of background noise in recordings. Existing models worked only on certain avian species and the development of further models required labeled data. The developed framework automatically extracted labeled data from available platforms for selected avian species. The labeled data were embedded into recordings, including environmental sounds and noise, and were used to train Convolutional Recurrent Neural Network (CRNN) models. The models were evaluated on unprocessed real world data recorded in urban KwaZulu-Natal habitats. The Adapted SED-CRNN model reached a F1 score of 0.73, demonstrating its efficiency under noisy, real-world conditions. The proposed approach to automatically extract labeled data for chosen avian species enables an easy adaption of PAM to other species and habitats for future conservation projects.
... Traditional methods of fauna monitoring often face limitations in terms of coverage and scalability (Prosekov et al., 2020). Manual observation is inherently constrained by human resources and time, making it difficult to gather data across vast territories or inaccessible regions (Sharma et al., 2023). Audio signal processing combined with machine learning techniques enables us to overcome these limitations. ...
... Audio signal processing offers a nonintrusive alternative to monitor these species without disturbing their natural habitats (Benocci et al., 2022). By analyzing acoustic recordings, we can obtain valuable insights into their vocalizations, migration patterns, and habitat preferences, all while minimizing our impact on their environment (Sharma et al., 2023;Stephenson et al., 2022). ...
The protection and monitoring of fauna species are essential for maintaining biodiversity and ensuring the sustainability of ecosystems. Traditional methods of fauna conservation and habitat monitoring rely heavily on manual observation and data collection, which can be time-consuming, and labor-intensive. In recent years, the application of machine learning techniques, such as object detection, has shown great potential in automating the identification of fauna species. In this study, we propose an approach to advancing fauna conservation through the utilization of machine learning-based spectrogram recognition. Specifically, we employ an object detection algorithm, YOLOv5, to detect and classify fauna species from spectrogram images obtained from acoustic recordings. The spectrograms provide a visual representation of audio signals, capturing distinct patterns and characteristics unique to different fauna species. Through extensive experimentation and evaluation, our approach achieved promising results, demonstrating a precision of 0.95, recall of 0.98, F1 score of 0.91, and mean Average Precision (mAP) of 0.934. These performance metrics indicate a high level of accuracy and reliability in fauna species detection. By automating the identification process, our approach provides a scalable solution for monitoring fauna populations over large geographical areas and enables the collection of comprehensive data, facilitating better decision-making and targeted conservation strategies.
This study investigates the capabilities of a tethered balloon system (TBS) for detecting and monitoring marine wildlife, primarily focusing on gray whales ( Eschrichtius robustus ) and various avian species. Over 55.7 h of aerial and surface footage were collected, yielding significant findings regarding the detection rates of marine mammals and seabirds. A total of 59 gray whale, 100 avian, and 6 indistinguishable marine mammal targets were identified by the airborne TBS, while surface-based observations recorded 1,409 gray whales, 1,342 avian targets, and several other marine mammals. When the airborne and surface cameras were operating simultaneously, 21% of airborne whale and 34% of airborne avian detections were captured with the airborne TBS camera and undetected with the surface-based camera. The TBS was most effective at altitudes between 50 to 200 m above ground, with variable-pitch scanning patterns providing superior detection of whale blows compared to fixed-pitch and loitering methods. Notably, instances of airborne detections not corroborated by surface observations underscore the benefits of combining aerial monitoring with traditional survey techniques. Additionally, the integration of machine-learning (ML) algorithms into video analysis enhances our capacity for processing large datasets, paving the way for real-time wildlife monitoring. Of the total number of blows detected by an ML algorithm, the percentage of blows identified by a human analyst was greater than that uniquely detected by the algorithm. Notably, more unique detections by the ML algorithm occurred during daylight, suggesting that sun artifacts may hinder human detection performance, thereby highlighting the added value of ML under these conditions. This research lays the groundwork for future studies in marine biodiversity monitoring, emphasizing the importance of innovative aerial surveillance technologies and advanced imaging methodologies in understanding species behavior and informing conservation strategies for sustainable marine energy, offshore wind development, and other marine resource management efforts.
