Finding asteroids is important in understanding the history of the solar system and uncovering threats of asteroid impacts. As we enter the age of large data output (TB to PB scales) from new surveys such as Euclid and Rubin, automated methods are needed to search for the large amount of asteroids hidden in the data. Furthermore, observations from current surveys such as OmegaCAM remain a valuable resource to improve the orbits of known asteroids and detecting new ones, complementing the observations from the aforementioned surveys. In this project, we develop a convolutional neural network (CNN) model that separates out asteroid streaks from other objects detected in OmegaCAM raw images. We will go through the data used to train and test the model, along with the challenges to guide the model towards accurate classification of asteroid streaks.
Keynote talk: Deep Learning for Image denoising – Jiayang Shi (CWI)
In this tutorial we take a hands-on tour of image denoising with (deep) learning. Starting from the classical BM3D baseline, we cover a broad range of modern approaches, supervised, self-supervised, and generative-model-based. We discuss what each assumes about the data and the noise. Accompanying Colab notebooks let you run these methods on a noisy face image in minutes, compare them side by side, and adapt them to your own denoising problem.
