8898 Seminar Earth Sciences
Speaker: Wei Ji Leong
Seminar Title: Teaching machines about our planet: Viewing, Learning, Imagining
To see how our planet is changing, and to be able to derive meaning from it quickly and automatically. Machine learning helps us make sense of Earth Observation data, but how to make sense of machine learning itself? From the natural landscape and built environment, to the vast oceans and depths of the Antarctic ice sheet, this talk will cover examples of various machine learning techniques used to extract insights from Earth Observation data. It starts from pixel-based classification of optical satellite imagery and Unet-based convolutional networks, going on to more advanced techniques like Reinforcement Learning and Generative Adversarial Networks. Beyond the pixel, we'll also look at clustering point clouds and using super-resolution to push the limits of spatial boundaries. At some point, you'll wonder where to focus your attention on in this field, and find out that Transformer models might be the next big thing to get into.
Host: Steven Lower
Zoom Meeting information
Click the Zoom link to join the Seminar
