In order to provide scientists with data critical for crafting effective conservation policies and approaches, the Gordon and Betty Moore Foundation and other partners, including Google, supported six conservation organizations to launch Wildlife Insights, a cloud-based platform that uses artificial intelligence to analyze images and information captured by “camera traps.”
Scientists routinely use camera traps — cameras equipped with infrared triggers — to gather data and capture images about wildlife to track its presence and behavior. Until now, however, the process of sorting these images has been painstakingly laborious and siloed. “We have camera traps and other passive sensors that collect information no matter what. You can put them in a very large area and they will snap a picture of an animal going by. The amount of information you can get is incredible,” explains Jorge Ahumada, senior wildlife conservation scientist at Conservation International and executive director of Wildlife Insights.
Now, with the AI-based platform, conservation experts like Ahumada can make better use of images of elusive anteaters, jaguars, and monkeys for example, to measure and study wildlife habitat, species and biodiversity. In addition to collecting and reviewing photos, the platform can provide analytics, including population trends — information that could help researchers understand whether a species is increasing its range or if its numbers are in decline.
The powerful new technology can classify images up to 3,000 times faster than humans can, analyzing 3.6 million photos in an hour. Once researchers upload the photos, Wildlife Insights automatically predicts which images include wildlife and the species captured. Wildlife researchers can then accept or correct the AI model’s prediction, which helps to continually improve the model’s accuracy.
“Wildlife Insights combines cutting-edge technological innovation with an unprecedented degree of cross-institution collaboration, giving the conservation community an entirely new level of visibility into the state of our natural world.” Aileen Lee, chief program officer for the Moore Foundation’s Environmental Conservation Program
The foundation has supported this partnership among Conservation International, Google, the Smithsonian’s National Zoo and Conservation Biology Institute, Wildlife Conservation Society, the North Carolina Museum of Natural Sciences, World Wide Fund for Nature, the Zoological Society of London and Yale University’s Map of Life.
To develop its identification capability, the platform began with nine million images submitted by the project’s six conservation partners. Any registered user— including citizens, scientists, teachers, and students—can now access the information and explore images from across the globe. Registration includes a vetting process to prevent access by poachers. Users can filter images by species, country and year, and download data instantly. As more researchers, wildlife conservation organizations, museums, scientific institutions, and private citizens join Wildlife Insights, the database will continue to grow and provide an increasingly rich, diverse and complex overview of global wildlife populations.
As camera trap expert Angélica Díaz-Pulido from Colombia’s Alexander von Humboldt Biological Resources Research Institute notes, “To the degree that we can obtain quality and timely information, we can create conservation measures that allow us to avoid large negative impacts and to try to conserve species before it’s too late.
Wildlife managers, for example, can gauge the health of individual species, while governments can use the data to inform policies. Organizations working to restore forests can ensure that wildlife is also returning to their habits. Indigenous communities that depend directly on wildlife for their survival can now monitor animal patterns.
“With a million species facing extinction, it's an all-hands-on-deck problem,” explains Tanya Birch, who leads Google Earth Outreach's Nature Conservation partnerships, like Wildlife Insights.
The foundation’s Environmental Conservation Program has supported Wildlife Insights within its special projects grantmaking, which has addressed a diverse array of issues through solutions that emerge from technological innovation and unprecedented partnerships.
What’s new with Wildlife Insights – Latest features as of Summer 2021
The latest Wildlife Insights release showcases features that bring data and conservationists together to protect wildlife.
Public downloads and data exploration
Wildlife Insights is open for any registered user to explore projects and download public data from the largest open-access database for camera trap data.
All data provided through Wildlife Insights is standardized and includes detailed information about how the data were collected. This means that exploring large scale patterns and comparing how populations change across time is easier than ever before. As more researchers and organizations join Wildlife Insights, the public database will continue to grow to provide a richer, more diverse snapshot of global wildlife populations.
With this release, there are multiple ways to explore the data:
- Search projects by moving around the map and clicking on any dot to view project details.
- Use basic and advanced filters to narrow your search by species, area, dates or project methodology.
Encouraging open data sharing
Collecting biodiversity data can be a costly endeavor. Data sharing encourages reuse, further extending the long-term conservation value of data. While many people agree that data sharing is important and are open to sharing, many remain hesitant to share without some assurance.
That’s why Wildlife Insights provides tools that encourage data sharing. With Wildlife Insights, data providers have control over when their data is shared and how it is shared.
- Embargoes give data providers and researchers the flexibility to keep their data private while they publish the data;
- Sensitive data including the exact locations of sensitive species (currently defined by IUCN’s categories) and images of humans are not shared publicly, unless the data provider approves access. These sensitive data remain visible only to the data provider.
- Citations are automatically generated for each project and are included in every download package and on the public project page. All citations include persistent identifiers (Archival Resource Keys, or ARKs), which provide a permanent reference to the project and ensure the long-term availability of the dataset. By providing a standard citation and ARK, Wildlife Insights helps to ensure that data providers receive credit and recognition for their data collection efforts.
Notifications to track a project’s impact
All data providers have access to a notifications center where they can manage their public data and track how it has been used. The Notifications center provides:
- Tools to approve or deny public requests for sensitive data. This gives data providers full control over their sensitive data.
- A message center. Registered Wildlife Insights users can contact any data provider to explore collaborations by sending a message through their public project page. All messages are displayed to the data provider in the Notification center.
- A summary of every public download request for the data provider’s projects. The information provided includes the requester’s email, the reason for the request and the expected product from the data download.
These tools provide transparency to the data provider in how their data is used. If a dataset is used for publication, the data provider or data user can reach out to discuss collaborations or co-authorship.
Data management improvements
The Wildlife Insights team has built many new data management features that make the processing phase easy, and let data providers focus on the important work.
- Bulk deployment uploads are a great way to collect standardized information while in the field and quickly enter the information in Wildlife Insights;
- Shortcut keys minimize effort needed to review and edit identifications on individual images;
- Bulk selections allow users to quickly select and identify multiple images at the same time;
- Subprojects (e.g., Protected area vs. Not protected area) allow users to group deployments so they can easily be analyzed across groups of interest;
- New roles - the Organization Contributor, Project Contributor, and Project Tagger - offer flexibility so that Organization and Project Owners can invite all their collaborators to join while still maintaining data privacy and data quality.
- Data and statistics are updated on demand, so the latest information is available anytime and anywhere.
Artificial intelligence model updates
The Wildlife Insights machine learning model is core to minimizing time spent processing data and maximizing the time spent analyzing and applying findings.
- Model has been trained by over nearly 12 million images to identify more than 732 species and 105 higher taxonomic classes;
- Bounding boxes are displayed when the model detects objects in images, making it easier to spot species of interest;
- A new model architecture increases accuracy 2% across the board.
- A taxonomic level rollup has been integrated so users see relevant results when the model isn’t confident at the species level
- A geofilter has been built into the model so that the model doesn’t predict a species that doesn’t occur in that country.
Stay tuned for more details on the upcoming Automated Analytics dashboard and an offline desktop client!
Message sent
Thank you for sharing.