Virginia Tech Launches AI-Powered Shark Monitoring Database.

Researchers at Virginia Tech, led by Francesco Ferretti, Assistant Professor in the Department of Fish and Wildlife Conservation, and collaborators at Stanford University, are developing sharkPulse, an open-access database leveraging artificial intelligence (AI) to autonomously extract data from online photographs. The methodology bypasses reliance on voluntary submissions by employing computational systems designed to scan online sources for shark imagery, automatically recording location data, timestamps, and species identifications. Validation of extracted data is achieved through a dual-verification process involving both public and expert review prior to inclusion in the searchable database, facilitating the mapping of shark populations and tracking of abundance and distribution changes at an unprecedented scale. Published in Fish and Fisheries on 11 August 2025, this approach transforms passive digital activity into valuable conservation data, addressing a critical knowledge gap hindering effective conservation efforts given the current at-risk status of approximately one-third of shark species and the limited understanding of their habitats and frequented areas.

Shark Data Deficiencies

A significant impediment to effective elasmobranch (shark, ray, and chimaera) conservation stems from pervasive data deficiencies concerning species distribution, abundance, and population trends. Despite approximately one-third of shark species facing extinction risk, comprehensive baseline data remains lacking for a substantial proportion of the global shark fauna. This scarcity hinders accurate assessments of conservation status, complicates the development of effective management strategies, and limits the ability to monitor the efficacy of existing protective measures.

Traditional methods of shark population monitoring, such as tagging, aerial surveys, and dedicated research expeditions, are often logistically challenging, resource-intensive, and geographically limited, resulting in incomplete spatial and temporal coverage. To address these limitations, a collaborative research effort involving Francesco Ferretti, Assistant Professor in the Department of Fish and Wildlife Conservation at Virginia Tech, and colleagues at Stanford University, is pioneering a novel approach utilising publicly available online photographic data. The initiative, culminating in a publication in Fish and Fisheries on 11 August 2025, centres on the development of ‘sharkPulse’, an open-access database leveraging artificial intelligence (AI) to systematically scan online sources for shark imagery.

This methodology circumvents the reliance on voluntary citizen science submissions, instead employing computational systems designed to perform tasks typically requiring human intelligence, specifically image recognition and metadata extraction. The system automatically harvests location data, timestamps, and preliminary species identifications from images, significantly increasing the volume and scope of data acquisition. The validation process is critical to ensuring data integrity.

Images identified by the AI are subject to a two-tiered verification system, involving both public review and expert confirmation. This hybrid approach combines the scalability of crowdsourcing with the taxonomic expertise necessary to resolve ambiguous identifications and minimise errors. Researchers emphasise that this shifts the paradigm of citizen science from relying on voluntary submissions to intelligent autonomous discovery, effectively transforming everyday digital activity into valuable conservation data.

The resulting database will facilitate the mapping of shark populations and the tracking of changes in abundance and distribution at an unprecedented scale and speed, providing crucial insights into species’ ecological requirements and habitat preferences. The ubiquity of digital cameras and the increasing prevalence of online image sharing represent a largely untapped resource for ecological monitoring. Encounters with marine megafauna, such as the sighting of a basking shark in Monterey Bay during whale watching excursions, can now contribute significantly to our understanding of shark populations. This approach is particularly valuable for documenting the distribution of elusive or infrequently observed species, and for tracking range shifts in response to environmental changes.

AI-Driven Discovery

The escalating threat to shark species – with approximately one-third facing extinction – necessitates innovative approaches to population monitoring and habitat assessment. Researchers at Virginia Tech, led by Francesco Ferretti, Assistant Professor in the Department of Fish and Wildlife Conservation, alongside collaborators at Stanford University and other institutions, are pioneering a novel methodology utilising artificial intelligence (AI) to construct sharkPulse, an expansive open database of shark sightings. This initiative addresses a critical data gap hindering effective conservation strategies by leveraging the vast and largely untapped resource of online photographic data.

The research, detailed in Fish and Fisheries on 11 August 2025, moves beyond traditional, volunteer-driven citizen science models. The core of the sharkPulse system lies in its automated image acquisition and analysis pipeline. Rather than relying on individuals to submit photographs, the platform employs AI algorithms – specifically convolutional neural networks – to scan online sources, including social media platforms and image repositories, for potential shark imagery.

These algorithms are trained to identify sharks based on visual characteristics, extracting crucial metadata such as geographical location (derived from geotags or contextual clues), timestamps, and preliminary species identifications. The system’s architecture is designed for scalability, enabling the processing of millions of images and the continuous expansion of the database. This automated approach significantly increases the volume and velocity of data acquisition compared to conventional methods.

A crucial aspect of the methodology is the implementation of a robust validation process to ensure data accuracy and reliability. Images identified by the AI are subjected to a two-tiered verification system. Initially, the public is invited to review and confirm the AI’s identifications, leveraging the principle of crowdsourcing.

Subsequently, expert taxonomists and marine biologists validate the public’s assessments, resolving ambiguous identifications and minimising potential errors. This hybrid approach combines the scalability of crowdsourcing with the taxonomic expertise necessary for robust data quality control. The validation process also incorporates confidence scores assigned by the AI, allowing researchers to prioritise images for expert review and assess the overall reliability of the data.

The resulting database, sharkPulse, offers unprecedented opportunities for mapping shark populations and tracking changes in abundance and distribution. By integrating data from diverse sources and employing advanced analytical techniques – including spatial modelling and time-series analysis – researchers can gain valuable insights into species’ ecological requirements, habitat preferences, and responses to environmental changes. This information is critical for informing conservation management strategies, such as the establishment of marine protected areas and the mitigation of threats to shark populations.

Furthermore, the methodology is applicable to the monitoring of other marine megafauna, offering a versatile tool for biodiversity conservation. The ubiquity of digital cameras and the increasing prevalence of online image sharing represent a paradigm shift in ecological monitoring, transforming everyday digital activity into valuable conservation data.

Conservation Implications

The development of sharkPulse, detailed in the 11 August 2025 publication in Fish and Fisheries, represents a significant advancement in the capacity for effective shark conservation. Currently, approximately one-third of all shark species face the threat of extinction, a crisis exacerbated by substantial gaps in fundamental knowledge regarding their distribution, abundance, and population trends. The innovative methodology employed by researchers – Francesco Ferretti of Virginia Tech, alongside collaborators at Stanford University and other institutions – directly addresses this data deficiency by harnessing the vast, readily available resource of online photographic data.

This shifts the paradigm of shark population monitoring from reliance on limited, often geographically biased, voluntary submissions to a system of intelligent autonomous discovery, significantly increasing both the scale and frequency of data acquisition. The implications for conservation management are substantial. Prior to platforms like sharkPulse, accurately mapping shark distributions and tracking changes in abundance required extensive and costly dedicated surveys, limiting the scope and temporal resolution of available data.

The ability to automatically extract location, timestamp, and species identification from publicly available images, coupled with a rigorous validation process involving both citizen scientists and expert taxonomists, allows for near real-time monitoring of shark populations across vast geographical areas. This capability is particularly crucial for understanding the impacts of climate change, overfishing, and habitat degradation on vulnerable shark species, informing the development of targeted conservation strategies. The validation methodology itself is a key strength of the project.

The two-tiered system – initial public review followed by expert confirmation – ensures a high degree of data accuracy and reliability. Expert validation resolves ambiguities and minimises errors, while the integration of AI-assigned confidence scores allows researchers to prioritise images for review, optimising the efficiency of the validation process. This approach not only enhances the quality of the data but also fosters public engagement in scientific research, promoting a greater understanding of shark ecology and conservation challenges. The resulting database facilitates informed decision-making regarding the establishment of marine protected areas, the implementation of fisheries management regulations, and the mitigation of threats to shark populations, ultimately contributing to the long-sustainability of these ecologically vital species.