yingkaisha/keras-unet-collection

how to use more than 3 channel input

rezadarooei opened this issue · 3 comments

rezadarooei commented

is it possible to use more than 3 channels as input? I want to concentrate on different images transforming and creating more than 3 channels. I used that method and put input_size=(512,512,4) but I got the error, is there any way I could use more than 3 channels?

image

yingkaisha commented

I will look into it around Friday.

  • ❤️1
yingkaisha commented

Hi, sorry for getting back to it late. I saw cuDNN launch failure in your error message. It is likely not related to the number of channels.

Would you mind sharing your code blocks? In addition, perhaps you need to track your GPU memory usage and Tensorflow/CUDA versions.

rezadarooei commented

hello dear,
I found the best way to solve this problem for more than 3 channel is using dataset type. it solves my problem.

  • 👍1