Satellite Earth observation has become a key technology in forest monitoring. With the European earth observation program Copernicus, a program exists that provides access to temporally high-resolution and freely available satellite images, and this worldwide. New analysis methods are needed to deal with the huge amounts of data; machine learning offers excellent opportunities here. Even with UAVs, huge amounts of data are often generated when, for example, ground resolutions of 2 cm are used. Concrete evaluation topics are e.g. the detection of game damage or the differentiation of tree species.
Topics which will be addressed:
- Introduction to Python and Tensorflow Keras
- Introduction to Deep Learning and neural networks
- Deep Learning architectures and their application fields
- Exemplary forestry applications
- Deep Learning with Python and Keras
In this module, students acquire key qualifications for the use of Deep Learning algorithms for forestry applications, which is also transferable to other applications in other disciplines. They learn the principles of Deep learning as well as neural networks and their optimisation. You will develop an understanding of which problems can be solved with the methods of Deep Learning and which methods should be selected. After completing the module, the students are able to freely program Deep Learning applications in Python. They can independently implement existing neural networks and handle implement existing neural networks and deal with large amounts of data.
The programme will be announced to the course participants after registration
The course is offered as a full-day block course. Part I on 21-22.10.2021 | Part II on 16-18.03.2022
Presentation (approx. 20 minutes, 20%) and project work (max. 10 pages, 80%) of the project groups.
- Dr. Nils Nölke (University of Göttingen)
- Maximilian Freudenberg, PhD student, University of Göttingen