Paper 13118-49
GedankenNet: self-supervised learning of holographic imaging enabled by physics consistency
22 August 2024 • 9:30 AM - 9:45 AM PDT | Conv. Ctr. Room 2
Abstract
We present GedankenNet, a self-supervised learning framework designed to eliminate reliance on experimental training data for holographic image reconstruction and phase retrieval. Analogous to thought (Gedanken) experiments in physics, the training of GedankenNet is guided by the consistency of physical laws governing holography without any experimental data or prior knowledge regarding the samples. When blindly tested on experimental data of various biological samples, GedankenNet performed very well and outperformed existing supervised models on external generalization. We further showed the robustness of GedankenNet to perturbations in the imaging hardware, including unknown changes in the imaging distance, pixel size and illumination wavelength.
Presenter
UCLA Samueli School of Engineering (United States)
Luzhe Huang is a Ph.D. candidate in the Department of Electrical and Computer Engineering at the University of California, Los Angeles. His research focuses on computational microscopy and computational imaging, with a particular emphasis on leveraging deep learning and artificial intelligence (AI) to solve intricate inverse problems across diverse imaging modalities. His work is dedicated to addressing critical limitations within AI applications in imaging, including challenges related to model generalization, reliability, and physical compatibility. Additionally, he is motivated to explore state-of-the-art computer vision and image processing technologies, such as generative AI.