In this module students are introduced to the concepts of machine learning and deep learning in order to process Remote sensing data. Remote sensing is the science that provides geometric and semantic information about objects at or near the surface of the Earth using the sensors which are installed on satellites or other airborne platforms. Along with fundamentals of remote sensing, some applications like object detection and classification especially on images and also regression algorithms on remote sensing observations will be covered.
In the context of image understanding, an introduction to digital image processing will be given, which deals with the application of filters on the images to extract the information which could be used in machine learning and deep learning algorithms.
Each of the lectures in this module is supplemented by practical parts to enable the students to process real-world remote sensing datasets, efficiently.
Prerequisites:
Students are expected to have a good knowledge on machine learning, calculus, linear algebra and probability theory. Moreover, at least basic knowledge of python programming, and image processing is required. Along with introducing the concepts of deep learning, the lectures will provide a refresher on relevant concepts from image processing and machine learning whenever required. However, familiarity with these concepts would be necessary (see lecture Machine Learning, offered in WS).
Content:
Details of deep learning architectures with a main focus of image-based problems will be discussed. The course will start with a refresher on machine learning and ANNs and during some first lectures the concepts of deep learning (DL) and convolutional neural networks (CNNs) will be presented. Implementing the well-known DL networks will also help the students to be prepared to tackle real world problems during their study and later in their carrier. The course would be a hands-on and project-based course. Hence, open-source and free libraries and frameworks form Python programming ecosystem will be used to implement various DL concepts on freely available datasets. Upon the completion of the final project of the course, students will acquire the skills to apply the learned concepts in a specific application area.
ECTS: 5 (Lecture+Lab session)
This Module consists of 2 courses
Location: Seminar room (Bienroder Weg, 81, 38106 Braunschweig)