Subprojects

Deep Learning

Deep Learning

The subproject of the Fraunhofer Institute is committed to the implementation of Deep Learning methods including applications for image data derived from ecological research. The strong structural form of artificial neural networks enables mechanical learning to identify target features and characteristics in images. Mechanical learning methods will significantly increase the efficiency of image feature identification through automated evaluation, enabling ecologists to process much greater digital image databases. Sample data for testing these methods will be supplied by biological research groups in the project that include a broad spectrum of data sources from pollen and roots, to bats, ensuring universal applicability in ecological sciences. The analyses (potentially supported by a PhD student) will focus on the automated detection of observationally defined objects in an image, incorporating automated segmentation to refine margins separating detected objects from background data. Next, isolated objects can be passed along to human experts or downstream software tools. Research tasks in the subproject are focused on “Deep Convolutional Neural Networks” given their proven suitability for evaluation image data. The method focus will shift gradually from “supervised” to “semi-supervised” until it reaches “unsupervised” classification. At this final stage, little to no data are used to classify prior to the start of the training. Additionally, the inclusion of semantical information (e.g. topology of root systems) will be assessed to enhance image analysis. This approach centers on improving the efficiency of image classification and processing, developing new fields of ecological research that relies on universally practicable methods.

Biomathematics

Biomathematics

Mareike Fischer manages the subproject “Biomathematics” with a PhD student and a PostDoc. Biomathematics mediates between mathematic theories and biological application, providing new tools to analyze biological data. M. Fischer will mediate between machine-learning-theoreticians and ecologists by incorporating biomathematical modeling and statistical evaluation. The subproject “Roots” will focus on modeling root growth temporally by using graph theory and explore if the level of root growth is statistically suitable to determine the nutrient distribution in the soil. Additionally, multidimensional data of bats, wood and pollen are further study targets for promoting deep learning methods in ecological research. Baseline data required to train machine learning approaches are numerous and include DNA (field of expertise for M. Fischer), sound recordings of bats, or the chemical properties of wood and pollen. DIG-IT!’s Deep Learning methods will identify data characteristics that are not visible to humans or can be derived from additional information.

Project coordination + wood anatomy

Project coordination + wood anatomy

Martin Wilmking leadsthe coordination of DIG-IT! and manages the subproject “Wood Anatomy”. This subproject will be supported by a PhD student as well as collaboration with partners from Fraunhofer IGD and associated master theses. The PostDoc “catalyst and multiplier” will link all members within the consortium (catalyst) and provides DIT-IT!-internal expertise for external users (multiplier). Synergisms with ongoing research projects (e.g. “WETSCAPES” project) and in the dendro-ecological laboratory “DendroGreif”. Assigned to the coordinator are two technical employees: a half position for supporting coordination and a full-position for assisting the users with data acquisition, digital image processing, and data management.

Roots

Roots

Jürgen Kreyling manages the subproject “Roots” to coordinate “automated digitalization of the dynamic in root systems of plants”. The subproject will establish the automated acquisition of various root characteristics (e.g. length, diameter) from mini-rhizotron scans in various settings including simplified laboratory conditions to complex field conditions. The project will include a PhD student and collaborations with partners from Fraunhofer IGD and associated master theses. J. Kreyling’s working group is equipped with a greenhouse and established research sites in a local beech forest and the nature reserve “Karrendorfer Wiesen”, enabling highly-resolved rhizontronscans acquisition over time. Further research collaborations with the “WETCAPES” project are expected.

Automated identification of species/individuals of selected animal species

Automated identification of species/individuals of selected animal species

Gerald Kerth manages the subproject “automated identification of species/individuals of selected animal species”. This project will include a PhD student and collaborations with partners from Fraunhofer IGD and associated master theses. Field data will exclusively support this project, with a special focus on 10 different wintering grounds of bats in Mecklenburg-Western Pomerania. The established methods for the digital determination of species will be applied at reasearch sites of the two DBU (Deutschen Bundestiftung Umwelt) projects "FUN" and "Living underground".

Pollen identification

Pollen identification

Hans Joosten manages the subproject “pollen identification” and will include a PhD student and collaborations with partners from Fraunhofer IGD, biomathematics (UG) and associated master theses. Laboratory analysis will be conducted predominantly in the laboratories of the working group Peatland Studies and Palaeoecology where specialized equipment, from boring, sample collection and preparation to pollen analysis, is available. For high-resolution images, microscopes from Imaging-Centre of the Biology Department of University Greifswald will be accessed. For coring beeches during flowering stage, the established transect of the working group J. Kreyling will be utilized. The sampling of stratified sediments will be carried out in cooperation with the German Research Centre for Geosciences (GFZ) in Potsdam.