Learn From Unpaired Data for Image Restoration: A Variational Bayes Approach

Zheng, Dihan; Zhang, Xiaowen; Ma, Kaisheng; Bao, Chenglong

doi:10.1109/tpami.2022.3215571

Cited by 8 publications

(2 citation statements)

References 65 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, in 3D classification, generating particles corresponding to less favored categories could potentially improve the process of particle classification. Moreover, the conditional generative model could be employed to synthesize paired training data for downstream tasks [41]. In signal recovery, a dataset composed of synthesized pairs of clean and noisy particles can be harnessed for training a denoising network, thereby enhancing the signal in the raw particles.…”

Section: Conclusion and Discussionmentioning

confidence: 99%

“…It is noted that the noise term n i contains both random error and the model error induced by the estimation error of parameters. Define Θ = {ψ, ϕ, V latent } to be the set of imaging parameters, the CVAE model learns a generative model that maximizes the conditional likelihood p ( x | Θ ), which is equivalent to the unpaired data modeling problem [41] between the latent projection v = C ( ψ ) P ( ϕ ) V latent and raw particles.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Addressing preferred orientation in single-particle cryo-EM through AI-generated auxiliary particles

Zhang,

Zheng,

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

The single-particle cryo-EM field faces the persistent challenge of preferred orientation, lacking general computational solutions. We introduce cryoPROS, an AI-based approach designed to address the above issue. By generating the auxiliary particles with a conditional deep generative model, cryoPROS addresses the intrinsic bias in orientation estimation for the observed particles. We effectively employed cryoPROS in the cryo-EM single particle analysis of the hemagglutinin trimer, showing the ability to restore the near-atomic resolution structure on non-tilt data. Moreover, the enhanced version named cryoPROS-MP significantly improves the resolution of the membrane protein NaXusing the no-tilted data that contains the effects of micelles. Compared to the classical approaches, cryoPROS does not need special experimental or image acquisition techniques, providing a purely computational yet effective solution for the preferred orientation problem. Finally, we conduct extensive experiments that establish the low risk of model bias and the high robustness of cryoPROS.

show abstract

Section: Conclusion and Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%