Camera-Trap Classification With Deep Learning Under Ground Truth Uncertainty

Deep learning can aid timely analysis of camera trap data to effectively monitor biodiversity responses to climate change. Image classifications collected through citizen science projects typically feature disagreement amongst volunteers, i.e., label ground truth uncertainty. We consider a combined camera trap and citizen science dataset featuring East African mammals and birds. We investigate (species-specific) test accuracy under different model choices and levels of training ground truth uncertainty on images with different levels of human classification difficulty.