Adaptive 3D descattering with a dynamic synthesis network

作     者：Waleed Tahir Hao Wang Lei Tian Waleed Tahir;Hao Wang;Lei Tian

作者机构：Department of Electrical and Computer EngineeringBoston UniversityBostonMA 02215USA Department of Biomedical EngineeringBoston UniversityBostonMA 02215USA 

出 版 物：《Light(Science & Applications)》 (光（科学与应用)（英文版）)

年 卷 期：2022年第11卷第5期

页      面：766-786页

核心收录：

学科分类：081203[工学-计算机应用技术] 08[工学] 0835[工学-软件工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：funding from National Science Foundation(1813848 and 1846784) 

主　　题：network scattering synthesis 

摘      要：Deep learning has been broadly applied to imaging in scattering applications.A common framework is to train a descattering network for image recovery by removing scattering *** achieve the best results on a broad spectrum of scattering conditions,individual“expertnetworks need to be trained for each ***,the expert’s performance sharply degrades when the testing condition differs from the *** alternative brute-force approach is to train a“generalistnetwork using data from diverse scattering *** generally requires a larger network to encapsulate the diversity in the data and a sufficiently large training set to avoid ***,we propose an adaptive learning framework,termed dynamic synthesis network(DSN),which dynamically adjusts the model weights and adapts to different scattering *** adaptability is achieved by a novel“mixture of expertsarchitecture that enables dynamically synthesizing a network by blending multiple experts using a gating *** demonstrate the DSN in holographic 3D particle imaging for a variety of scattering *** show in simulation that our DSN provides generalization across a continuum of scattering *** addition,we show that by training the DSN entirely on simulated data,the network can generalize to experiments and achieve robust 3D *** expect the same concept can find many other applications,such as denoising and imaging in scattering ***,our dynamic synthesis framework opens up a new paradigm for designing highly adaptive deep learning and computational imaging techniques.