From elephants raiding crops in villages, lions killing livestock and humans to herders retaliating by killing lions, human-wildlife conflict continues to be a growing concern not just in Kenya but across the world.

Human and wildlife casualties, impoverished communities due to damage to crops and loss of livestock are some of the profound implications of the conflict. Across the country, human-wildlife conflict cases processed by the County Wildlife Conservation and Compensation Committees have risen to about 16,000. These claims are estimated to be worth Ksh. 4.5 billion.

Further, with 1071 deaths occurring between 2014 and 2020, and 218 between 2022 and 2023, there is an urgent need for measures to control human-wildlife conflict for the communities involved and the wildlife at stake.

To enable peaceful co-existence and resource-sharing between humans and wildlife, Vodafone is currently testing a technology solution that is running as a prototype and will help secure community safety and livelihoods. The solution, dubbed M-Twiga, with the help of cameras and solar power, uses artificial intelligence (AI) and the Internet of Things (IoT) to detect wildlife at a distance.

M-Twiga is Amy Turner’s brainchild, inspired by her experience in East Africa. Amy, a Vodafone project Creator and Lead, saw firsthand how elephants destroyed water infrastructure.

“I’d also heard stories of lion predation on livestock pens in East Africa, specifically in Kenya, and I think both of those experiences and hearing people’s first-person distress at this situation made me realise that we need to do something,” she says.

Once the concept received funding to get started, Amy collaborated with Joe Griffin, Vodafone’s Senior Sustainability Manager, with whom they co-led the project.

The idea was developed in a Vodafone lab, and in March 2024, the prototype finally left the lab and entered the Mara Siena Conservancy in Kenya, where testing began.

“We developed two prototypes and took them to Mara Siena Conservancy in March to test the AI, the cameras, the hardware, and the software system,” says Joe, adding, “We put one of the prototypes on top of a vehicle so that we could go out and collect images and data of four animals we were interested in: elephants, lions, leopards, and hyenas. We also left one unit static in a wetland area within the Mara Siena ecosystem.”

From the test, it was clear that the hardware – the cameras, batteries, and solar panels – was working optimally. But what they really wanted to see was whether the AI model they had built could collect data and correctly identify animals.

The team had previously tested the AI model at a wildlife park in the United Kingdom, but unfortunately, it did not yield the results they were hoping for.

Joe explains that most object detection AI models for animals seek to identify an animal that’s shoulder height. With M-Twiga, the cameras are positioned three metres high to see into the middle distance so that they can identify, alert and deter the presence of a predator early.  “If you look at an animal from three metres high, it looks quite different than if you look at it along. And we’ve got some images that show this because we tested on some hyenas in the UK. When the AI was looking at a hyena from three metres high, it was actually identified as a baboon.”

“I know it sounds ridiculous, but when you look at the image of a hyena from three metres high, there are similarities. It’s very broad and has a thick upper body, and the neck and the limbs are foreshortened because of the angle. So, it’s quite easy to trick AI. I mean, AI is not human intelligence yet,” Joe says.

During the testing phase in Kenya, over 50,000 data points were collected from a camera the Vodafone team left at the Mara Siena Conservancy, under the guardianship of Mara Siena wardens. The images will be used to refine the AI and redesign the hardware.

M-Twiga also takes into account the needs of the communities that live near wildlife populations. From Amy’s experience, many communities use dogs to alert them to predators. But now, M-Twiga, through IoT connectivity, will send an SMS alert to the Kenya Wildlife Service (KWS) and community leaders or farmers. The text message, which will work on a 2G network for areas with low connectivity, will provide details of the type of animal, the area or direction it’s located, and the level of certainty.

The technology is designed to sit in remote areas as well as in the heart of communities. When they roll it out, the idea is to set it up on top of a three-metre pole at the centre of a livestock boma or shamba, looking out to 25 to 30-metre parameters. M-Twiga might just be the welcome, groundbreaking solution to curbing the mounting human-wildlife conflict cases.