Hello. I got some problem when I run the Cuda_SRCNN for big size images.

[star4s]

Hello. I got some problem when I run the Cuda_SRCNN for big size images.

When I try running the Cuda_SRCNN for 654 * 664 size image, the neural network for Super_resolution works well.

But when I try running for big size images, 1120 * 1450, I got the fail message. The message is

vl::impl::nnconv_forward_blas: getWorkspace [out of memory error] .

Do you know why I got the error messege for big size images?

In my case, the 654 * 664 size image is good to work in the Super resoultion.

Thank you for your attention

Best.

[layumi]

This mean your GPU is out of memory. You may try net.conserveMemory=true; first.

[star4s]

Thank you for your kindness. I got it. As you know, the commend is already added net.conserveMemory=true; But it is not still worked. Do you know how I can solve the problem? My graphic card is Geforce GTX 980. The graphic card should be upgraded? Thank you for your help.

[layumi]

GPU Memory matters. If you have enough money, 1080TI(11GB) is the best card for the time being. Or you can try 1080(8GB).

[star4s]

thank you for your Kindness. I got it. I will try using the 1080TI(11GB).