Rising numbers of wolf (Canis lupus) populations make traditional, resource-intensive methods of wolf monitoring increasingly challenging and often insufficient. This study explores how wolf howls can be used as a new monitoring tool for wolves by applying Artificial Intelligence (AI) methods to detect and classify wolf howls from acoustic recordings, thereby improving the effectiveness of wolf population monitoring. Three AI approaches are evaluated: BirdNET, Yellowstone’s Cry-Wolf project system, and BioLingual. Data were collected using Song Meter SM4 (SM4) audio recorders in a known wolf territory in Klelund Dyrehave, Denmark, and manually validated to establish a ground truth of 260 wolf howls. Results demonstrate that while AI solutions currently do not achieve the complete precision or overall accuracy of expert manual analysis, they offer tremendous efficiency gains, significantly reducing processing time. BirdNET achieved the highest recall at 78.5% (204/260 howls detected), though with a low precision of 0.007 (resulting in 28,773 false positives). BioLingual detected 61.5% of howls (160/260) with 0.005 precision (30,163 false positives), and Cry-Wolf detected 59.6% of howls (155/260) with 0.005 precision (30,099 false positives). Crucially, a combined approach utilizing all three models achieved a 96.2% recall (250/260 howls detected). This suggests that while AI solutions primarily function as powerful human-aided data reduction tools rather than fully autonomous detectors, they represent a valuable, scalable, and non-invasive complement to traditional methods in wolf research and conservation, making large-scale monitoring more feasible.
Welfare assessment for the endangered red panda (Ailurus fulgens) in captivity requires systematic behaviour monitoring, yet traditional direct observation is often limited by observer subjectivity and time constraints. This study evaluates the feasibility of employing machine learning (ML) to automate behavioural monitoring of a red panda in a complex, mixed-species enclosure at Aalborg Zoo, Denmark. Using video data from cameras in the enclosure of the red panda, and the machine learning model LabGym for animal detection and behavioural categorisation, models were trained to analyse activity patterns of the red panda. The results demonstrate that while the behaviour categorizer is a promising tool with high classification confidence, the overall system effectiveness is currently limited by the object detector’s performance in a naturalistic environment. Challenges such as environmental obstructions such as rocks, foliage, and trees, and the animal’s camouflage contributed to a significant amount of unclassified time, which may affect the overall assessment of behavioural distribution. We conclude that while ML holds potential for non-invasive behaviour monitoring, its application in complex zoo settings requires improved detection capabilities to be fully reliable. Future iterations of this system could be enhanced by complementing standard object detection with pose estimation frameworks. Implementing alternative labelling strategies or background subtraction methods could additionally mitigate the detection challenges posed by environmental obstruction.
Welfare assessment for the endangered red panda (Ailurus fulgens) in captivity requires systematic behaviour monitoring, yet traditional direct observation is often limited by observer subjectivity and time constraints. This study evaluates the feasibility of employing machine learning (ML) to automate behavioural monitoring of a red panda in a complex, mixed-species enclosure at Aalborg Zoo, Denmark. Using video data from cameras in the enclosure of the red panda, and the ML model LabGym for animal detection and behavioural categorisation, models were trained to analyse activity patterns of the red panda. The results demonstrate that, while the behaviour categorizer is a promising tool with high classification confidence, the overall system effectiveness is currently limited by the object detector’s performance in a naturalistic environment. Challenges such as environmental obstructions (e.g., rocks, foliage, and trees) and the animal’s camouflage contributed to a significant amount of unclassified time, which may affect the overall assessment of behavioural distribution. We conclude that, while ML holds potential for non-invasive behaviour monitoring, its application in complex zoo settings requires improved detection capabilities to be fully reliable. Future iterations of this system could be enhanced by complementing standard object detection with pose estimation frameworks. Implementing alternative labelling strategies or background subtraction methods could additionally mitigate the detection challenges posed by environmental obstruction.
Rising numbers of wolf populations make traditional, resource-intensive methods of wolf monitoring increasingly challenging and often insufficient. This study explores how wolf howls can be used as a new monitoring tool fro wolves by applying AI methods to detect and classify wolf howls from acoustic recordings, thereby improving the effectiveness of wolf population monitoring. Three AI approaches are evaluated: BirdNET, Yellowstone's Cry-Wolf project system, and BioLingual. Data were collected using SM4 audio recorders in a known wolf territory in Klelund Dyrehave, Denmark, and manually validated to establish a ground truth of 260 wolf howls. Results demonstrate that while AI solutions currently do not achieve the complete precision or overall accuracy of expert manual analysis, they offer tremendous efficiency gains, significantly reducing processing time. BirdNET achieved the highest recall at 78.5% (204/260 howls detected), though with a low precision of 0.007 (resulting in 28,773 false positives). BioLingual detected 61.5% of howls (160/260) with 0.005 precision (30,163 false positives), and Cry-Wolf detected 59.6% of howls (155/260) with 0.005 precision (30,099 false positives). Crucially, a combined ap-proach utilizing all three models achieved a 96.2% recall (250/260 howls detected). This suggests that while AI solutions primarily function as powerful human-aided data re-duction tools rather than fully autonomous detectors, they represent a valuable, scalable, and non-invasive complement to traditional methods in wolf research and conservation, making large-scale monitoring more feasible.