Skip to main content

Stay Connected to Dartmouth on:
Facebook Facebook
YouTube YouTube
Flickr  Flickr
Twitter Twitter
Instagram Instagram
Google+ Google+

 

Higher Education Act Information

 

Office of Communications
7 Lebanon St., Suite 201
Hanover, N.H. 03755
Phone: (603) 646-3661
communications@dartmouth.edu
Home >

New Dartmouth Software Tracks Wildlife with Photos, Not Tranquilizers

Takeaways:

  • New wildlife image-recognition software proves nearly 100 percent accurate for tracking giraffes, meaning they no longer have to be captured and tagged.
  • Scientists now have image algorithms for species with irregular, spotted and striped markings.
  • Growing number of scientists using image algorithms to study a global menagerie of wildlife.
  • Software is more accurate, less invasive, less expensive, less time consuming and covers more territory than traditional mark-recapture and aerial survey methods.

Feb. 3, 2015

Wildlife scientists are typically pictured getting up close and personal with their subjects, capturing and tagging them to better understand how they live, where they go and other mysteries. But a growing number of researchers worldwide are taking a hands-off approach, using image-recognition software instead to identify and track species with unique markings. Algorithms have been developed for cheetahs and other species with spotted patterns and for zebras and other creatures with striped markings. Now, Dartmouth scientists have developed a remarkably accurate software for giraffes and other species whose coat or skin patterns are irregular in size, shape and edges.

The “Wild-ID” program has proven so accurate – with an error rate of virtually zero – on giraffes that Dartmouth researchers no longer need to capture them as they conduct the largest ever study of the iconic ungulates and one of the largest ever individually-based demography studies of a large mammal.

The software also is drawing interest from scientists around the world who study a range of other species with irregular patterns, including wildebeest, salamanders, lynx, anacondas, terrapins, an African antelope species, turtles, a giant Mongolian salmon species, trout, chameleons, butterflies, toads and others. The software is available free to researchers.

Wildlife pattern-recognition algorithms – whether for irregular, spotted or striped markings that are as unique as human fingerprints – are designed to be more accurate, less invasive, less expensive, less time consuming and cover more territory than traditional mark-recapture and aerial survey methods in demographic studies, which include a species’ population size, birth and death rates and other characteristics. The programs are designed for large population species such as giraffes and aren’t applicable to tigers and other species whose numbers are so small that it’s still manageable to compare their images manually in hard copy photo catalogues.

Dartmouth and other researchers have published studies using Wild-ID with giraffes in Tanzania in 2012,  wildebeest migration in 2012 and 2014 in Tanzania, and giraffe social relationships in Namibia and endangered salamanders in Texas in 2013. Dartmouth researchers are using the program for a new study of giraffes in Uganda. PDFs of the studies are available on request.

Wild-ID is part of Dartmouth’s partnership with the Giraffe Conservation Foundation to study giraffes’ abundance, reproduction, survival and movements. Giraffes, who live only in sub-Saharan Africa, have declined in population about 40 percent over the past decade to an estimated 80,000 because of habitat fragmentation and bushmeat poaching. But they remain understudied compared to other iconic species and are not listed as threatened because of a dearth of information about their ecology and demography.

Photographic mark–recapture with still images has mostly been used with small populations of marine mammals and mammalian land predators, but it has been less effective for large wildlife populations because as the number of images grows, the hard copy photo catalogues become unwieldy and the probability of error increases. In more recent years, there have been efforts to use computers to semi-automate the matching process, including image analysis algorithms to extract, store and match pattern information from digital photo and video images.

Since the Dartmouth team started using Wild-ID in 2011, their database has grown to about 800 individual giraffes, or about 65 percent of Tanzania’s Tarangire Ecosystem population, far more than could be captured and identified without the aid of a pattern-recognition program. Tanzania has the largest population of giraffes in the world. The software has reduced tracking costs by about 100-fold for giraffes and more than 50-fold for wildebeests compared to conventional mark-recapture methods.

Available to comment are:

Last Updated: 9/9/15