Island Conservation uses Lenovo and NVIDIA machine learning solutions to restore island ecosystems

Optimizing and accelerating image processing with AI will help conservation experts protect seabird nesting sites on Robinson Crusoe Island.

Biodiversity around the world is under threat. We are currently in what many scientists call the Sixth Mass Extinction, with hundreds of species of plants and animals having been permanently lost over the past century.

Island ecosystems can be particularly vulnerable to human activities. On Robinson Crusoe Island in the South Pacific, native seabirds such as the great-footed shearwater are easy prey for the invasive South American coati. Introduced to the island by humans nearly a century ago, coatis are domestic cat-sized mammals in the same family as raccoons that hunt shearwaters at nesting sites across the island.

Protecting the island's ecosystem

Leading the fight against invasive species on Robinson Crusoe Island is Island Conservation, an international non-profit organization that restores the island's ecosystem to benefit wildlife, oceans and local communities. Island Conservation has been working with island residents for many years to protect endangered and endangered species.

Physically removing invasive coatis from petrel nesting sites is only part of the challenge for island conservation. To track coati activity, the organization also uses more than 70 remote camera traps to closely monitor petrel nesting sites.

Process thousands of images every month

The organization's camera traps generate a large amount of data, approximately 140,000 images each month. These data need to be collected and analyzed for signs of coati activity. Previously, island conservation teams relied heavily on manual processes for this task. Classifying 10,000 images takes approximately 8 hours of non-stop work by trained experts.

Additionally, manual processing diverted valuable resources from important island protection work on the ground. The organization knew there had to be a better way.

Realize the potential of machine learning

David Will, Head of Innovation at Island Conservation, recalls the challenge: “We started experimenting with machine learning. [ML] A model that speeds up image processing. We believed automation was the way to go, but one of the big challenges was connectivity. Many of the ML solutions we considered required moving all photos to the cloud for processing. But Robinson Crusoe Island didn't have a reliable enough internet connection to do that. ”

As a temporary workaround, Island Conservation stored camera trap images on SD cards and sent them by airmail to Santiago de Chile so they could be uploaded to the cloud for processing. Airmail was the fastest and most frequent means of transport between the island and the mainland, but the service ran only once every two weeks. Additionally, there were delays of up to three months between when the camera traps captured images and when island conservation groups received analysis.

David Will commented: “The time between detecting an invasive species with cameras and being able to respond means there was not enough time to make the decisions needed to prevent extinction on the island.”

Tackling infrastructure challenges

That's where Lenovo stepped in. A global team of 16 volunteers visited the island, funded by the Lenovo Work for Humankind initiative, which has a mission to use technology for good. Using Lenovo's smarter technology, from devices to software and IT services to servers, volunteers will help upgrade the island's network infrastructure, increasing bandwidth from 1 Mbps to 200 Mbps. I was able to carry out my daily tasks while doing so.

Robinson Crusoe Island suffers from harsh marine conditions and access is restricted. They needed a robust system that would bring computing to their data and enable remote management. The solution was his ThinkEdge SE450 from Lenovo with an NVIDIA® A40 GPU. AI-optimized edge servers offer a rugged design that can withstand extreme conditions while running quietly, allowing you to operate comfortably in your new remote workspace. Lenovo worked with Island Conservation to tailor the server to its needs, adding graphics cards to increase his AI processing power per node. “We took the supercomputer capabilities they had in Santiago and brought it to a much smaller form factor,” said Charles Ferland, vice president and general manager of edge computing at Lenovo. I am.

ThinkEdge SE450 eliminates the need for on-site technicians. Unlike data centers, which require on-site staff, ThinkEdge servers can be monitored and serviced remotely by members of his Lenovo team. It proved to be the perfect solution. ThinkEdge servers enable complete remote access and management of devices, reducing decision-making from months to days.

David Will commented: “Lenovo allows us to run both A40s at the same time, which greatly increases processing speed. This was not possible before.” Almost everything is done on the ThinkEdge SE450. ”

Unleash the power of automation

To automate both coati detection and classification, Lenovo data scientists at the AI ​​Center of Excellence detect and separate results from coati and other species from MegaDetector, an open-source object detection model that identifies animals. I built a custom AI script. People and vehicles that appear in camera trap images. Lenovo data scientists then trained an ML model on the custom dataset to provide multiclass classification results for his nine species on Robinson Crusoe Island, including petrels and coatis.

This two-step GPU-accelerated detection and classifier pipeline can provide results for 24,000 camera trap images in just 1 minute. Previously, for a trained professional to do this he needed 20 hours of labor. This saves you an amazing 99.9% of your time. The model achieved 97.5% accuracy on a test dataset with approximately 400 classifications per second. By leveraging the power of NVIDIA's CUDA-enabled GPUs, we were able to speed up MegaDetector by 160x compared to previous implementations.

Sachin Gopal Wani, AI Data Scientist at Lenovo commented: “Providing solutions that are easy for users to interpret is a key part of our AI leadership. We created a custom script that produces compatible output, allowing even non-technical end users to visualize much faster without having to save additional images. 's solution can load the overlay of the original image and classification results, saving terabytes of disk space.

Using these ML capabilities, Island Conservation can filter out images that do not contain invasive species with a high degree of certainty. The organization uses a newly upgraded internet connection to upload images of coati activity to the cloud, where volunteers on the mainland can assess images and quickly send recommendations to the islands.

“ML can speed up image processing, deliver results in minutes, and reduce strategic decision-making time from three months to weeks,” says David Will. “This reduced response time means more birds are protected from direct predation and populations recover faster.”

Looking to the future

Island Conservation plans to continue collaborating with the Lenovo AI Center of Excellence to develop Gen AI to detect other types of invasive species, including rodents, another major threat to native animals. .

“Lenovo's support shows us how easy it is to train models to detect other invasive species on Robinson Crusoe Island,” said David Will. “We recently set up a test environment to detect new species. After just 7 hours of training the model, we recorded a detection accuracy of 98%, which is an impressive result.”

As the scope of the project expands, Island Conservation plans to further use Lenovo ThinkEdge SE450 devices with NVIDIA A40 GPUs for new projects across other islands. Lenovo's ThinkEdge portfolio is optimized for edge AI inference, delivering superior performance and robustness to process data securely where it's created.

With the help of Lenovo and NVIDIA technology, Island Conservation is in a stronger position than ever to protect native species from invasive threats.

David Will said: “In many of our projects, more than 30% of the total project cost is spent removing the last 1% of intruders and making sure they don't exist. Using Lenovo By being able to make decisions based on hard data rather than intuition, Island Conservation can tackle new projects faster.”

heal our oceans

Island Conservation's work with Lenovo on Robinson Crusoe Island serves as a blueprint for future work. The research team plans to reuse the AI ​​application to detect a variety of invasive species on different islands around the world, from the Caribbean to the South Pacific, Western Pacific, Central Indian Ocean, and Eastern Tropical Pacific. I am. This aims to protect endangered species and increase the number of endangered species. Improving biodiversity and climate resilience.

In fact, Island Conservation, Re:wild, and the Scripps Institution of Oceanography recently brought together NGOs, governments, funders, island communities, and individuals to create 40 globally significant islands and oceans by 2030. Launched the Island-Ocean Connection Challenge to begin comprehensive restoration of the island's ecosystem. .

“Everything is interconnected in what is known as the land-sea cycle,” says David Will. “Healthy oceans depend on healthy islands. Island and ocean ecosystem components circulate with each other and share essential nutrients for the plants and animals within them. Indigenous cultures have believed in this for centuries. Climate change, ocean degradation, invasive species, and biodiversity loss are disrupting entire land and sea ecosystems and disproportionately impacting island communities.”

The Island-Ocean Connection Challenge marks the beginning of a new era in conservation that breaks down artificial silos and focuses on holistic restoration.

David Will concludes: “Our joint efforts, supported by Lenovo and NVIDIA, will help bridge the digital divide for island communities, enabling them to use cutting-edge technology to restore, rewild, and restore ecosystems. “Advances in AI will leave them further behind.”

JOIN NOW www.jointheiocc.org.

Click here to read Lenovo's story on island conservation. Or click here to watch Lenovo's case study video.