Ce diaporama a bien été signalé.
Le téléchargement de votre SlideShare est en cours. ×

Unsupervised deep learning methods for biological image reconstruction and enhancement

Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Prochain SlideShare
Abstract
Abstract
Chargement dans…3
×

Consultez-les par la suite

1 sur 13 Publicité

Plus De Contenu Connexe

Plus récents (20)

Publicité

Unsupervised deep learning methods for biological image reconstruction and enhancement

  1. 1. Unsupervised Deep Learning Methods for Biological Image Reconstruction and Enhancement Mehmet Akckaya, Burhaneddin Yaman, Hyungjin Chung, Jong Chul Ye Hyungjin Chung
  2. 2. Introduction
  3. 3. Introduction: Inverse Problems Inverse problems 𝒚 = 𝐻 𝒙 + 𝒘 Regularized Least Squares (RLS) 𝑥 = arg min 𝒙 ‖ ‖ 𝒚 − 𝐻 𝒙 2 2 + 𝑅(𝒙)
  4. 4. Self-supervised learning based methods
  5. 5. Self-supervised learning based methods Batson et al. 2019 Denoising min 𝜽𝑑 𝑛=1 𝑁 𝐹𝜽𝑑 𝒚𝑛 − 𝒚𝑛 2
  6. 6. Self-supervised learning based methods Reconstruction min 𝜽𝑟 1 𝑁 𝑛=1 𝑁 ℒ(𝒚Θc 𝑛 , 𝐻Θc 𝑛 (𝐹𝜽𝑟 𝒚Θ 𝑛 , 𝐻Θ 𝑛 )) Yaman et al. 2020
  7. 7. Self-supervised learning based methods: Results Denoising Yaman et al. 2020 Batson et al. 2019 Reconstruction
  8. 8. Generative model based methods
  9. 9. Generative model based methods Geometry of generative models
  10. 10. Generative model based methods: VAE • Vanilla VAE (Kingma et al. 2014) • Spatial-VAE (Bepler et al. 2019) • DIVNOISING (Prakash et al. 2021)
  11. 11. Generative model based methods: GAN
  12. 12. Generative model based methods: cycleGAN Geometry of cycleGAN
  13. 13. Generative model based methods: Results ODT reconstruction Chung et al. 2021

×