PiLocNet: Physics-informed neural network on 3D localization with rotating point spread function

Description

We consider the 3D localization problem of the point spread function (PSF) engineering and propose a novel framework based on the physics-informed neural network (PINN), namely PiLocNet, to solve this problem.

Our PiLocNet combines deep learning and variational methods, which enhances the black box neural networks by employing the known physics information of the forward process into the framework as the data fitting term. In the meantime, it incorporates the regularization terms from the variational method that best fits the noise model. This work focuses on the single-lope PSF, while it is widely applicable to other PSFs or other imaging problems.

Period	20 Jun 2025
Event title	SFB Conference: Tomography Across the Scales
Event type	Conference
Location	Strobl, AustriaShow on map
Degree of Recognition	International