Using automated recognition software to identify leopards from their spot patterns
After the few seconds spent in awe following spotting a leopard, we usually grab our camera to record that rare encounter. As scientists, we do the same, but with an additional goal in mind: we want to know who this leopard is! We can recognise individual leopards by their unique rosettes pattern and we have been using camera traps deployed at waterholes to estimate our leopard population size. Until now, each picture was attributed to an individual by comparing it manually to a database of individual pictures. Needless to say, this process is highly time consuming‼ With the advances in computer vision technology over the past few years, it is now possible to use software that do that job for us. We have recently been checking the capabilities of one called HotSpotter and the results are quite amazing. It takes but a few seconds for the software to identify a coat pattern, compare them to all of the other images in the database and present us with a set of the 10 most likely matches for us to choose from. And more often than not, the 1st match is a correct one. The computer can do in seconds what can take us hours! Stay tuned for more automated image recognition news when we start to experiment with other species such as zebras, giraffes and spotted hyaenas!