Since 2017, over 20,000 volunteers have classified hidden-camera photos from Kenya; more than 4,000 showed giraffes, including 175 calves

For seven years, tens of thousands of ordinary internet users clicked through blurred, motion-triggered trail-camera photos of dusty ground, empty grassland, and the occasional goat, marking most of them “nothing interesting” before moving to the next image. According to the Zooniverse results page for the Wildwatch Kenya project, more than 20,000 volunteers took part between June 2017 and the project’s recent conclusion, classifying over one million images from motion-triggered trail cameras in Northern Kenya. What sounds like an unremarkable task turned into a genuinely useful dataset for conservation, with its own strengths and limitations.

The giraffe numbers

Per the Zooniverse results page, volunteers classified over 4,000 giraffe images, including exactly 175 showing calves. That calf count matters because the reticulated giraffe, the subspecies that dominates this part of Kenya, is in steep decline. According to the Giraffe Conservation Foundation, the species now numbers around 20,900 in the wild, down from an estimated 36,000 roughly three decades ago, a drop of more than 40%. The reticulated giraffe was listed as Endangered on the IUCN Red List in 2018.


Livestock monitoring was part of the plan

The most frequently classified animals weren’t giraffes at all, but livestock: camels, goats, sheep, cattle, and donkeys, along with domesticated dogs, which together accounted for more than 20,000 classifications. That wasn’t an accident the researchers simply tolerated: tracking livestock alongside wildlife was central to the project’s design, since, according to the Zooniverse results page, around 95% of reticulated giraffe habitat lies outside officially protected areas.

Comparing giraffe sightings to livestock density in those shared landscapes helps scientists gauge whether giraffes are being crowded out. The dataset also turned up more than 40 other species, including servals, wildcats, wild dogs, leopards, and lions, and has been used to update both giraffe and hyena range maps.

Can untrained volunteers be trusted with this?

This is a valid question to ask. Crowdsourcing can be a bit of a stigma. A 2020 study titled “Camera settings and biome influence the accuracy of citizen science approaches to camera trap image classification,” published in the journal Ecology and Evolution, compared three Zooniverse camera-trap projects. Each image was reviewed by multiple volunteers, their answers were combined into a single best answer, and a sample was then checked against expert classifications.

Wildwatch Kenya reached 83.4% accuracy, compared with 97.9% for Snapshot Serengeti and 93.5% for AmazonCam Tambopata. The study also found that images incorrectly labeled as empty were more likely to come from open grassland (27%) than from woodland (10%). Researchers attributed Wildwatch Kenya’s comparatively lower accuracy to camera setup rather than volunteer error, and recommended three-shot burst cameras, a narrower field of view, and field-tested sensitivity settings for future projects. Several of the study’s authors were affiliated with the San Diego Zoo Institute for Conservation Research, which has since been renamed the San Diego Zoo Wildlife Alliance following a 2021 rebrand.

What it adds up to

The images collected have also been shared with communities in Kenya, many of whom live alongside these animals but rarely see them directly. Wildwatch Kenya’s classification phase is now listed as inactive on Zooniverse, with no further data currently being collected, though its archive remains part of a broader catalog of Zooniverse projects spanning biology, climate, and space that may be worth a look for anyone curious about citizen science. The project didn’t single-handedly rescue the reticulated giraffe. But it does show that a clear process, a large volunteer base, and rigorous error-checking can turn a repetitive task into data that holds up to scrutiny.