Iterative human and automated identification of wildlife images

Camera trapping is increasingly used to monitor wildlife, but this technology typically requires extensive data annotation. Recently, deep learning has significantly advanced automatic wildlife recognition. However, current methods are hampered by a dependence on large static sets when intrinsically dynamic and involves long-tailed distributions. These two drawbacks can be overcome through hybrid combination of machine humans in the loop. Our proposed iterative human automated identification approach capable from imagery with distribution. Additionally, it includes self-updating that facilitates capturing community dynamics rapidly changing natural systems. Extensive experiments show our achieve ~90% accuracy employing only ~20% annotations existing approaches. synergistic collaboration machines transforms relatively inefficient post-annotation tool collaborative on-going annotation vastly relieves burden enables efficient constant model updates.