WISDAM (Wildlife Image Survey – Detection and Mapping): Software to produce data suitable for abundance estimation and spatial modelling from aerial photographic surveys of marine and terrestrial megafauna

Advancements in the capabilities of imaging technology and aerial platforms (e.g. drones, satellites) have enhanced our capacity to survey wildlife abundance, distribution and habitat use, and have improved the accuracy of the data collected. We developed our WISDAM (Wildlife Image Survey – Detection and Mapping) open-source software to support aerial imagery surveys by standardising and automating the process of extracting data from the images. The software is based on our experience developing imagery/drone survey methods and our intrinsic understanding of the data requirements for wildlife surveys. Although many researchers are interested in automating the review of the images, there has been little attention on the manual review of images, a process critical for training new Artificial Intelligence (AI) models, improving existing models, and providing reliable data for species where model performance may be insufficient, particularly for species occurring at very low densities. A standardised manual review process is critical for maintaining a consistent probability of detecting animals throughout an image survey dataset. The current capabilities of WISDAM include manually labelling objects (e.g. detected animals) with metadata, mapping these detections along with aerial image footprints (including over marine landscapes) to real-world coordinates, labelling images with environmental attributes and other metadata, identifying multiple detections of individual animals (manually or according to spatial referencing), automatic matching of detections from multiple sources (e.g. multiple manual reviewers or AI models), export of data as CSV files or in multiple GIS formats, and export of images and detections to dedicated folders for training AI models. WISDAM also allows the import and verification of detections from AI models. We encourage contributions (e.g. recommended additions, bug fixes, and AI training imagery) from users to enhance the tool's capabilities. WISDAM is already in use by a several research and local community groups in numerous countries.