Deep learning for terahertz image denoising in nondestructive historical document analysis

HIGHLIGHTS

• who: Balaka Dutta from the Institut University Hospital have published the paper: Deep learning for terahertz image denoising in nondestructive historical document analysis, in the Journal: Scientific Reports Scientific Reports
• what: To overcome the lack of paired data, the authors propose to apply an unsupervised learning network, in particular CycleGAN, to generate paired images using unpaired synthetic clean images and real noisy images. The overall objective function is, *, G *=arg min GAB GAB,GBA max DA,DB LGAN (GAB, DB ) + LGAN (GBA, DA ) + cyc Lcyc (GAB, GBA ).
• how: One dataset was made with . . .
  

   

  

  If you want to have access to all the content you need to log in!

  Thanks :)

  Username or E-mail:

  If you don't have an account, you can create one here.