Machine Vision
Machine Vision for Natural History
The ongoing climate and biodiversity crises require biodiversity and ecological knowledge based on data at relevant long-temporal and broad-spatial scales – this data can be derived from natural history collections. Norway has a rich history of natural history collections, dating back to the 18th Century. A substantial quantity of the natural history specimens held in Norway’s museums are digitized, detailing the taxon sampled, date and location published online. The NTNU University Museum has over one million natural history objects dating from the 1700s to present, many of these, in particular the vascular plant herbarium, are already photographed.
The Machine Vision for Natural History research group focusses on using computer vision and machine learning to extract ecological and taxonomic data from images of natural history specimens, including plants and invertebrates. This includes data on the timing of flowering, specimen size and species’ traits including leaf area and wingspan. We develop and train models to classify specimens and segment and measure key traits. By applying these models over large numbers of specimens collected over long periods and wide spatial extents, we can test how species have changed over time and in response to global environmental changes.