AI Paper Atlas

Research map / Image generation / Image restoration

LWGNet – Learned Wirtinger Gradients for Fourier Ptychographic Phase Retrieval

Atreyee Saha, Salman S. Khan, Sagar Sehrawat, Sanjana S. Prabhu, Shanti Bhattacharya, Kaushik Mitra

ECCV 2022 · Image generation › Image restoration · approach: Inverse imaging with learned priors · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use complex-valued operations instead of real-valued operations on real and imaginary channels (ablation)LWGNet · simulated dataaverage PSNR: increases the average PSNR by ~5dBImproved
process stochastic gradients with the learned network instead of enhancing the object field after Wirtinger updates (ablation)LWGNet · simulated datareconstruction qualityImproved
use stochastic gradients instead of regular gradients (ablation)LWGNet · simulated datareconstruction qualityImproved
increase unrolled stages from 3 to 5 for 8-bit data (ablation)LWGNet · simulated 8-bitPSNR: 3 stages -> 5 stagesImproved
increase unrolled stages beyond 3 for 16-bit and 12-bit data (ablation)LWGNet · simulated 16-bit and 12-bitPSNR: saturates after 3 stagesNo clear effect
use LWGNet instead of F-cGAN at 8-bitLWGNet · Simulated FPM histology (8-bit)PSNR: 25.715 -> 26.035Improved
use LWGNet instead of F-cGAN at 12-bitLWGNet · Simulated FPM histology (12-bit)PSNR: 26.325 -> 27.726Improved
use LWGNet instead of F-cGAN at 16-bitLWGNet · Simulated FPM histology (16-bit)PSNR: 29.021 -> 28.745Hurt
use LWGNet instead of F-cGAN at 16-bitLWGNet · Simulated FPM histology (16-bit)SSIM: 0.907 -> 0.829Hurt
use LWGNet instead of F-cGAN in parametersLWGNet · Simulated FPM histologyparameters (millions): 7.88 -> 0.39Improved
use LWGNet instead of F-cGAN in inference timeLWGNet · Simulated FPM histologyinference time (ms): 29 -> 341Hurt
use naive unrolled Wirtinger flow with convolution block (ablation)naive unrolled WF · simulated dataconvergenceHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy