1.2 million sets of images from camera traps in the Serengeti National Park have been catalogued and this dataset is described online in Scientific Data. Out of these image sets, 322,653 contained animals, with 40 separate species identified, including animals such as the aardwolf, zorilla and the honey badger.
Camera traps have been in use for a number of years and have played an important role in wildlife conservation, allowing researchers to observe the presence of rare species in remote or protected areas. Advances in technology have expanded the capacity of camera traps while reducing their cost. This has resulted in a dramatic increase in these studies and an overwhelming increase in the data produced, creating a need for new ways to process images.
Alexandra Swanson and colleagues worked with a citizen science platform to develop the website where image sets from 225 camera traps were made available to the public. Volunteers then classified each image, identified species, counted the number of individuals and characterized their behaviours. The dataset covers images taken between 2010 and 2013 and were catalogued by more than 28,000 registered volunteers. The authors then designed an algorithm to seek consensus in the classifications and determine what was in the image. They hope that their resulting dataset may be used for further ecological research and for education purposes, and that their algorithm may contribute to future crowdsourced image processing projects.