Séminaire IPAG

---

Estimating differential pistons for the Extremely Large Telescope using deep learning

jeudi 16 octobre 2025 - 11h00

Pierre Janin-Potiron - LAM
--- 
Differential piston between the ELT’s M4 segments arises from multiple sources and degrades image quality, much like the low wind effect on the VLT, making its correction essential for achieving optimal performance. In this presentation, we show how differential piston can be estimated from focal plane images using a deep learning model based on a ResNet architecture. Using input images created from numerical simulations, we demonstrate the robustness of this method against multiple parameters, including seeing, detector noise, and chromatism, and derive the ideal operating range for this sensor. These results highlight the potential of AI for wavefront sensing and open new opportunities for wavefront correction on the ELT.

---

Hôtes : Alexis Carlotti

Salle Manuel Forestini, 414 rue de la piscine, 38400 Saint Martin d'Hères