Artificial intelligence helps protect endangered species

The mission of the nonprofit WPS (Wildlife Conservation Program) is to use technology to protect endangered species and ecosystems. To this end, the wpsWatch platform is being built using artificial intelligence on remote camera images. Support the installation of more cameras in critical wildlife habitats around the world and expand their work from identifying threats, classifying species and assisting with anti-poaching to preventing human-wildlife conflicts.

The organization captures more than 25,000 photos from cameras every day. “No one can look at every image and immediately understand what’s in it, but it’s critical to our operations,” said Matt Hron, director of product and accounts at Wildlife Solutions. “That’s what AI can do. arrive.

Beyond the scope of this image analysis work, the organization is expanding the use of the technology into other application areas. It requires the ability to quickly build and deploy new AI models to meet the varying needs of these efforts.

The organization's wpsWatch platform analyzes and monitors massive amounts of images from remote cameras located at more than 100 sites in nearly 20 regions. It is powered by Microsoft Azure VMs (virtual machines) and NVIDIA GPUs (graphics processing units) and is initially focused on security and anti-poaching goals within the organization's mission. ​

To this end, WPS worked with the Microsoft AI for Earth team to provide images for MegaDetector, an AI model developed by the AI ​​for Earth team to accelerate the detection of surveillance camera images. deal with. It’s a mutually beneficial relationship, with WPS using MegaDetector to help enhance and refine its wpsWatch monitoring solution, which provides image injection that contributes to the continuous improvement of the model. WPS provides its services and platform to protected areas free of charge.

Use MegaDetector, a computer vision anomaly detection model, to detect animals, people, and vehicles in your camera. It is designed in a way that supports the organization's desire to achieve new goals. “Because it is a standards-based application, many of our field users are able to choose the right hardware based on their specific needs,” said WPS executive director Eric Schmidt. “It gives us the flexibility to use a variety of smart way to adapt third-party systems so that we can work with those systems to make them more like a closed ecosystem."

MegaDetector A new version (V5) was released last year, and WPS saw improvements in accuracy immediately after implementing the new version.

One performance metric of the wpsWatch platform is the time it takes from receiving an image to identifying what triggered the image. Once the image is received, it takes seconds to obtain the AI ​​inference data to understand what is in the photo. MegaDetector v5 runs on the infrastructure and image analysis runs 50% to 60% faster than before. Specifically, the average processing time using MegaDetector v4 is about 2 seconds. The average time for new versions is between 500 and 700 milliseconds. "This is a big improvement, especially as we dramatically increase the number of images analyzed," said James Goodheart, software developer at WPS. Another enhancement using v5 is to improve accuracy. "We put out some older images that may not have been detected or were flagged for retraining. Some have been successfully detected in newer versions of the AI," Goodheart said.

In addition to image analysis, WPS uses other Microsoft infrastructure elements in its platform. For example, when the remote camera starts up, image data is emailed from the remote camera via the SendGrid service and then parsed using the WPS API. (The time required to transfer images varies depending on available communication services. Most use local mobile services, while some cameras connect over Wi-Fi.)

Photos use Microsoft Azure Blob storage , the metadata is delivered to WPS by Microsoft SQL Server. The photos are then forwarded to various AI image recognition solutions to determine what is in the photo, such as a vehicle, a person or an animal species of interest, which can then alert relevant teams on site based on what is in the image.

Expand into new domains

One area where WPS hopes to leverage its AI infrastructure is to support its efforts in preventing human-wildlife conflicts. This requires the ability to look for species in images and then recognize that, for example, elephants may be traveling along corridors toward human settlements, where they may be damaging crops. Or look for lions or wolves approaching livestock areas and alert locals to take precautions.

Additionally, WPS is conducting more detections of invasive species. What is needed is the ability to monitor rats, cats, dogs, goats or any local invasive species, combined with appropriate methods to ensure that there are no further invasions in the area. In each case, WPS uses the same app and camera to find the threat, whether it's a human or wildlife.

WPS wants to do all it can to encourage people to engage with global wildlife issues. “One of the really exciting things is that people around the world are able to directly participate in international wildlife conservation. With the tools we provide, anyone can monitor our data as a volunteer and become a contributor to poaching incidents and wildlife conservation around the world. First Responders to Animal Crime." Through cloud-based technology and data flows, individuals can have a global impact. Everyone can make an impact for nature conservation around the world.

The above is the detailed content of Artificial intelligence helps protect endangered species. For more information, please follow other related articles on the PHP Chinese website!