top of page
BPC GREY LOGO.jpg
CC GREY LOGO.jpg
COEX GREY LOGO.jpg

OVER 5 MILLION SNAPSHOTS COVERING 3,000 KM² , 24/7!

During 2023, together with the University of Zurich, we continued to monitor and research seasonal variations and long-term population trends of both predator and herbivore species using camera traps. Despite a few cameras having to be retired due to mechanical damage we were still able to cover the majority of BPC’s core historic 3,000 km2 study area.


To sort the enormous number of pictures collected (literally millions!) and optimize a workflow that would allow us to quicky process them, we teamed up with an engineering MSc student from EPFL, Raphaël Destriau. The workflow developed is as follows: In a first step, pictures are run through the Megadetector, an open-source tool that allows separating pictures containing animals, pictures containing vehicles/humans, and pictures only containing vegetation (false trigger). In a second step, pictures containing animals are run through Traptagger, an additional open-source tool that automatically assigns to each image the species photographed. As no AI tool is perfect, some species sometimes get mistaken, as in the case of the lion and their smaller cousin the caracal, or the cheetah and the leopard. Therefore, in a final step, these “troublesome” species had to be manually cross-checked. After manual labelling, we obtained >8,000 pictures of hyenas, >3,300 pictures of lions, 2,600 pictures of leopards and 1,500 pictures of African wild dogs collected over a couple hundred independent detection events. This means that, on average, we had dogs passing in front of one of our cameras every 2-3 days only!


This information can now be used to investigate patterns of habitat use, abundance, and distribution for carnivores but also for herbivore species, of which impala, with > 100,000 pictures, is by far the most detected species. Furthermore, images can be passed through the African Carnivore Wildbook (see next page) to try to identify individuals, although this may be challenging, especially for

low-quality, black-and-white pictures taken at night.

bottom of page