De-noising HST images with U-Nets

Screenshot 2021-02-24 at 17.23.10

We are applying U-Nets (a type of deep learning neural network) to improve the Signal to Noise on HST/ACS and HST/WFC3 images by modelling the noise and subtracting it from the images afterwards. After a successful first investigation (Vojteková et al. 2020), we are now planning to apply the algorithm to the whole eHST archive and to analyse the robustness of the solution to enhance better science with the same data and released the de-noised images as High Level Science Product. This will be followed up by an exciting review of interesting objects of our scientific interest. 😉

More info in our Machine Learning group website.