I'd like to use GANs for a 1-D image-to-image transformation

30 visualizaciones (últimos 30 días)
Ram Krishnan
Ram Krishnan el 7 de Nov. de 2024 a las 21:11
Comentada: Ram Krishnan el 8 de Nov. de 2024 a las 13:56
I'm working on a medical application where I have 1-D images (think ECG, PPG, etc). I'd like to use a GAN to mimic another 1-D image in a one-to-one way. One way to do transfer learning from a 2-D Cycle GAN example shown in Matlab is to replicate my 1-D image (125 X 1) into a 125 X 125 to form an image but that seems like a huge computing burden. Any ideas?
Thanks
  2 comentarios
William Rose
William Rose el 8 de Nov. de 2024 a las 4:41
125x125 is not a hge computing burden.
Goodfellow et al. (2014) use images for their examples, in the original GAN paper.
Zhu et al. (2017) use images, not 1-D signals.
Zhao et al. (2017) use images, not 1-D signals, in their paper.
I think you will need to do some original thinking. I would start with thinking about a 1-D convolution on a signal, instead of a 2-D convolution on an image, for the generator.
Good luck with your research.
Ram Krishnan
Ram Krishnan el 8 de Nov. de 2024 a las 13:56
Thanks for your advice, William. As you say, a 1-D convolution may be necessary instead trying a transfer learning from 2-D image networks.

Iniciar sesión para comentar.

Respuestas (0)

Categorías

Más información sobre Image Data Workflows en Help Center y File Exchange.

Productos


Versión

R2024b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by