Poaching is a major problem for wildlife preserves. While these organisations are struggling to protect animals on the verge of extinction, poachers do whatever they can to sneak in and kill them for immediate gains, usually something as stupid as an animal¡¯s coat, horn or other body parts.
Now, researchers from the University of Southern California Center for Artificial Intelligence in Society are using AI to spot these illegal hunters in near-real time and catch them before they can do harm.
The thing is, infrared cameras are great for picking up poachers when they strike in the night, but they can¡¯t distinguish between the heat signatures of animals and humans. What a team of USC reporters did is they employed a piece of proprietary technology to tag and label 180,000 heat maps of people and animals captured by the cameras. Using this data, they then trained a deep learning algorithm to recognise when it was a human the camera was seeing. They called it ¡®SPOT¡¯, the Systematic POacher deTector.
The researchers then put the technology to the test catching poachers at national parks in Zimbabwe and Malawi, with the AI processing live feeds from infrared cameras recording from drones flying above. Unfortunately, the AI took about 10 seconds to process images, far too long a period when a poacher might be sneaking up on an innocent rhinoceros, elephant, or tiger. But after much tweaking, the algorithm was able to spot poachers or animals in a much shorter time, about three-tenths of a second.
When put into service, SPOT could help ease the infrastructural burden on wildlife preserves using drones to prevent poaching. After all, instead of having people poring over the feeds constantly, you can just have a computer detect intrusions in almost real-time.
Maybe then all the endangered species can finally have a chance at revival.