Speaker
Description
Herbarium specimens offer valuable insights into changes of plant distribution from the past to the present, aiding in predicting future developments and responses to environmental changes. Over centuries, herbarium specimens have been collected, labeled with data like date and place of collection, collector, species name and information on habitat, and archived manually, leading to gigantic collections of important historic information. However, extracting information relevant for research, like phenological information, from such collections manually is a considerable challenge, which is why significant digitization efforts are taking place to enable faster processing. Unfortunately, although digitized herbarium specimens contain large amounts of phenology-relevant visual and textual information, automated approaches to extract phenological data from specimen images are still very limited.
We introduce a novel tool based on artificial neural networks to automatically extract phenological information from images of herbarium specimens. With this method, it will be possible to extract phenological information for large numbers of herbarium specimens without manual effort at high throughput and speed, which will enable further research on plant phenology from historical to present observations. Moreover, we develop the tool to be applicable to many different plant species without the need to supply own data or re-train the model on novel species.
The system is developed in interdisciplinary cooperation between biologists and computer scientists.
| Status Group | Postdoctoral Researcher |
|---|---|
| FOR TALKS: Poster Presentation Option | Yes, I’m willing to present as a poster. |