AI in The Wild: Leveraging AI for Conservation

Happy World Wildlife Day! We love reading about how engineers and researchers are using AI and machine learning to help protect vulnerable animals and ecosystems. Many of these tech-led conservation efforts rely on computer vision to collect the valuable data that is needed for researchers to make a significant impact. 

 Given that AI-driven solutions are our specialty, we thought it would be appropo to highlight some computer vision based conservation efforts that we found especially cool. 

Here’s an inspiring story of WildTrack’s efforts to protect the black rhino population in Zimbabwe and Namibia from poachers. Using non-invasive footprint monitoring, called Footprint Identification Technology (FIT), researchers analyze images of footprints using width, length, and even the unique “fingerprints” on the pads of the rhino’s feet–to identify the weight and sex of individual animals with up to 99% accuracy. This technology allows conservationists to track specific animals’ comings and goings, making monitoring and protecting easier.

Image Courtesy of WildTrack’s Website

Microsoft’s Zamba Cloud is another massive conservation undertaking that relies on computer vision to identify and catalog more than 20 species of animals. Hidden field cameras record hours of footage of animals in their natural habitats allowing for greater insight into habits and behavior. Traditionally, the video footage is broken into image frames where individual animals are then labeled using data annotation tools, like those we provide with our toolset, Sixgill Sense, to create an image dataset. The dataset is then used by engineers to train a machine learning model to identify individual animals. These deep learning models help protect vulnerable species from harm and are used to monitor and detect invasive species that have devastating consequences to local ecosystems.

One such use case could be the recently publicized wild pigs in Texas and other parts of the Southern US. Wild pig species cause $2.1 billion dollars worth of damage every year. A study from 2019 estimated that wild pigs were directly responsible for $118.8 million in agricultural damage in Texas alone. Not to mention the large scale environmental disturbance caused by the animal’s rooting behavior which can lead to increased soil erosion and changes in ecological chemical composition altering natural ecosystems and wetlands. It’s a big problem for the region.

While corral traps with motion activated cameras have proven to be an effective method for population control they still require human monitoring–between 2 to 5 hours of work per wild pig caught–which translate to man hours for landowners and state officials working to curb population growth. A possible solution would be to incorporate a computer vision machine learning model that is able recognize when the feral pigs enter the corral and then automatically close the gate behind them. Sixgill happens to be experts when it comes to using computer vision to identify hogs, so, if anyone is interested in pursuing this…just sayin’.

AI, machine learning, and computer vision continue to open up new possibilities when it comes to conservation efforts around the world. We can’t wait to see what additional ecological use cases will blossom in the future.

More Plainsight Blog Posts: