Subaru Seminars

Ground-Based direct imaging of exoplanets is limited by atmospheric turbulence, which introduces phase errors resulting in complex "speckle" patterns forming in the Point Spread Function (PSF). Understanding the relationship between the wavefront sensor (WFS) telemetry and the speckles is essential for high-contrast imaging; however traditional analytical reconstructions often struggle with the non-linearities in the WFS response. An alternative is a machine learning-based approach using a U-Net architecture. By treating PSF reconstruction as an image-to-image translation task, the model is trained to map telemetry from the wavefront sensor directly to the corresponding focal plane intensity distribution. This approach aims to leverage the U-Net’s ability to capture spatial features across multiple scales, potentially offering a more robust method for handling atmospheric speckles compared to standard linear models. This presentation outlines the current status of the model architecture, training data, and current model performance.

TBD

TBD

TBD

TBD

TBD