AI can now recognize wild animals in camera-trap images

Artificial intelligence will come to play an increasing role in wildlife conservation. If you need proof of that statement, here comes news of a new computer model developed by researchers at the University of Wyoming that can accurately identify and classify images of wild animals from photographs taken by motion-activated cameras.

Images captured by so-called camera traps placed in biodiverse locations can provide invaluable information about local wildlife populations. Yet these cameras can take millions of images that need to be analyzed, usually one by one. People doing that can be a tiresome and time-consuming process.

So the researchers have taught computers to do that on their own. “We trained machine learning models using convolutional neural networks with the ResNet-18 architecture and 3,367,383 images to automatically classify wildlife species from camera trap images obtained from five states across the United States,” they explain in a newly published study.

The computer model was also tested on an independent subset of 5,900 images of moose, cattle, elk and wild pigs from Canada. It produced an accuracy rate of 81.8%. In addition, during a test with images taken in Tanzania, the model was 94% successful in removing “empty” images, those without any animals. That means that human researchers themselves would have to inspect only the remaining 6% of images, which would make their task a whole lot easier.

Yet the software did even better with animals in the US. “The trained model classified approximately 2,000 images per minute on a laptop computer with 16 gigabytes of RAM,” the researchers note. “[It] achieved 98% accuracy at identifying species in the United States, the highest accuracy of such a model to date.”

This breakthrough in artificial intelligence, when applied to wildlife conservation on a large scale, could be a significant aid in the study and conservation of wildlife around the planet. The newly developed computer model is available in a software package for Program R, a commonly used free software employed for statistical computing.

“The ability to rapidly identify millions of images from camera traps can fundamentally change the way ecologists design and implement wildlife studies,” the researchers stress.