- Scientists have deployed artificial intelligence models to identify and re-identify endangered giraffes in Tanzania.
- The Wild Nature Institute partnered with Microsoft’s AI For Good Lab to launch Project GIRAFFE which uses open-source AI tools to identify and re-identify individual giraffes based on spot patterns on their bodies.
- The data has helped scientists come up with estimates on survival and reproduction rates, movements, and behavior of the animals.
Monica Bond and Derek Lee’s “love and obsession” for giraffes started during a trip to Uganda in 2005. Since then, they said, their mission was clear: return to Africa and lead the “world’s greatest giraffe study.”
Five years later, the duo co-founded the nonprofit Wild Nature Institute, based in Tanzania, where the Masai giraffe (Giraffa tippelskirchi) is the national animal. Here, their team has focused on identifying individual giraffes based on the unique patterns on their bodies. For years, the team did the work manually, after which they experimented with different iterations of technology. Neither approach yielded the results they were looking for, however.
Now, artificial intelligence models — developed in collaboration with Microsoft’s AI For Good Lab — are helping the team speed up their work. Project GIRAFFE (Generalized Image-based Re-identification using AI for Fauna Feature Extraction) is an AI-based, open-source tool that’s helping Bond and Lee efficiently identify and re-identify individual giraffes.
“It can now be done in minutes, and we can have the output the same day we collect the data,” Lee told Mongabay in a video interview.
Identifying individual giraffes, and being able to subsequently re-identify them, is crucial to understanding the areas that are important for their survival. It’s an urgent task, given how giraffe populations have declined in the last few decades primarily due to habitat loss and poaching. Moreover, protecting giraffe habitats serve a larger purpose for the ecosystem.
“Giraffes take up a lot of space themselves,” Lee said. “If you can protect enough habitats to have a healthy giraffe population, there’s lots of other plants and animals that are also benefiting from that same protection.”
But identifying and re-identifying individual animals from patterns on their bodies is no easy feat. Previously, scientists would copy the patterns in their notebooks in order to be able to re-identify animals on later expeditions. “As you can imagine, you can only identify so many animals that way with the human eye,” Bond told Mongabay in a video interview.
As part of their organization’s Masai Giraffe Project, Bond and Lee’s team would drive around to spot giraffes and capture photos of patterns on the right side of their bodies. But with six expeditions organized annually year after year, the team has gathered millions of photos over time.
Their earlier methods required the team to manually crop photos to focus on the giraffe’s body, a process that took them weeks. A separate model was then deployed to identify individuals. However, as photos from expeditions piled up, the processing time increased drastically. “It started to kind of fail,” Bond said. “Every time we put new data in, it would take longer and longer, like a week or two just for matching. That was hampering our research.”
The use of artificial intelligence models has now helped speed up the process. The methodology involves multiple algorithms that crop giraffes out of images, identify the chest area for patterns, and then compare the extracted patterns with the ones in the data set built by the Wild Nature Institute.
“The task of the algorithm is to compare that picture with other giraffes,” Juan Lavista Ferres, chief data scientist at the AI for Good Lab, told Mongabay in a video interview. “And see which of these giraffes have the same pattern.”
The technology has helped the team at the Wild Nature Institute to draw inferences about entire populations across huge swaths of the ecosystem. “Every time we see a giraffe again, we are then able to match the photos and build an encounter history,” Lee said.
The data have helped them come up with estimates for survival and reproduction rates, movements, and behavior of the animals. These, in turn, have proved critical to making targeted decisions on conservation and habitat management. Bond and Lee said they’ve also been able to understand the multilevel social structure of giraffes across landscapes. “And even within one landscape, we have been able to understand [in] which particular parts of the landscape the giraffes are doing well and where they are doing poorly,” Bond said.
The duo said they plan to continue to use the data and technology to identify critical areas for the animals, while also working with local communities to ensure peaceful human-animal coexistence.
“A big part of our mission is to see that people and the giraffes are both thriving together,” Lee said. “We see it on the ground already where humans and wildlife coexist, and it’s just the matter of will and willpower.”
Banner image: Endangered reticulated giraffes in East Africa. Giraffe populations have declined in the last few decades primarily due to habitat loss and poaching. Image by Rhett A. Butler/Mongabay.
Abhishyant Kidangoor is a staff writer at Mongabay. Find him on 𝕏 @AbhishyantPK.
Where are the giraffes hiding? Predictive tracking tech points the way
Credits
This story first appeared on Mongabay
This article is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License.
You may republish this article, so long as you credit the authors and Mongabay, and do not change the text. Please include a link back to the original article.
