Vendi information gain for active learning and its application to ecology

While monitoring biodiversity through camera traps has become an important endeavour for ecological research, identifying species in the captured image data remains a major bottleneck due to limited labelling resources. Active learning—a machine learning paradigm that selects the most informative data to label and train a predictive model—offers a promising solution, but typically focuses on uncertainty in the individual predictions without considering uncertainty across the entire dataset. We study Vendi information gain (VIG) as an active learning policy that selects images based on their impact on dataset‐wide prediction uncertainty, capturing both informativeness and diversity. We applied VIG to two representative datasets in ecological image classification, the Kenya Camera Trap Dataset and the Missouri Camera Trap dataset, and compared VIG against common active learning methods. On both datasets, VIG achieves higher accuracy with fewer labelled examples. We further conducted extensive ablation studies to investigate the effects of the predictive model, batch sizes and other hyperparameters to find relatively robust performance from VIG. VIG has broad applicability beyond ecology, and our results highlight its value for biodiversity monitoring in data‐limited environments.