The second method looks at the feature groups of individual lion faces, akin to how we identify faces through reduction; even if there is a mud splatter on a friend’s face we recognize the shape of their eyes. This method leverages a fine-tuned Deep Convolutional Neural Network with Residual Inception Blocks (“Inception ResNet v2”) pretrained on ImageNet and fine-tuned on the normalized LINC datasets labelled by conservationists. This network architecture is computationally efficient as it exhibits a high accuracy to parameter count ratio and it converges quickly on the LINC dataset when fine-tuned on a pre-trained ImageNet model.
Both these AI methodologies in conjunction with the continued integration of human-centered development of the user interface, allow researchers to handle and access previously unmanageably large datasets. The LINC project is an open source model allowing a sustained development that benefits the whole conservation community.
Learn More about the technology >